Painting a picture on analysis strategies for the photodegradation products of dyes

(1)

MSc Chemistry

Analytical Sciences

Literature Thesis

Painting a picture on analysis strategies for the

photodegradation products of dyes

by

Gerben van Henten

12481181

9 July 2020

Van ’t Hof Institute for Molecular Sciences

Supervisor:

Examiners:

Iris Groeneveld Prof. dr. Maarten van Bommel

(2)

2

Abstract

In the field of cultural heritage, there is a strong need for a better understanding of the light-induced degradation of dyes for the identification, restoration, and preservation of art. Elucidating the photodegradation mechanism is challenging as it is often difficult to establish the relationship between the starting material and the degradation products. This is further complicated by external parameters affecting the photodegradation such as the pH, solvent, substrate, humidity, and oxygen content and multiple degradation pathways occurring at the same time. It is necessary to address these parameters simultaneously with mechanism elucidation in controlled environments to develop online-degradation tools and increase the understanding of photodegradation. Various techniques have been applied in photodegradation studies such as chromatographic, spectroscopic and mass spectrometric applications. It is essential to know the type if degradation products that are generally being formed as there is no single system or technique capable of measuring all degradation products. LC-MS is the most versatile technique for dye analysis and currently the primary choice for photodegradation studies often supported by spectroscopic methods. However, recent developments in ambient MS techniques and SERS offer alternatives for simplistic mock-ups studies or when investigating a specific degradation mechanism. This review provides an overview of the different parameters affecting photodegradation, the techniques applied in photodegradation studies, and discusses the experimental design, through the photodegradation of triarylmethane and flavonoid dyes.

(3)

3

1. Introduction

Many dyes are known to degrade when exposed to light for prolonged periods of time. The influence of light irradiation induces reactions altering the chromophore of the dye, resulting in a diminished or complete loss of color. This can be beneficial, for example, in the removal of dyes from wastewater and the forensic analysis of documents such as authenticity and age determination, but is often an undesirable phenomenon. Photodegradation is particularly troublesome in the field of cultural heritage where the fading of colors affects the sustainability and aesthetical value of objects of art. A better understanding of light-induced degradation is fundamental for the identification, restoration, and preservation of cultural heritage objects. Elucidating the photodegradation mechanism is affected by numerous parameters such as pH, temperature, wavelength of irradiation, humidity and the substrate, which all influence the photodegradation kinetics. Most photodegradation studies tend to focus on a limited amount of parameters resulting in assertions that are not universally applicable for different dyestuffs. The current procedure to measure individual colorants at different conditions is tedious and time-consuming. However, the alternative to perform degradation studies on samples containing multiple compounds, is challenging as it is often difficult to establish the relationship between the degradation products and the starting materials. Developing new strategies to track photofading and analyze the photodegradation products and intermediates is essential to deconvolute complex samples and elucidate the photodegradation mechanisms.

Research on dye analysis and their photodegradation products has been performed by a variety of chromatographic (e.g. liquid chromatography (LC), gas chromatography (GC)), spectroscopic (e.g. Infrared (IR), Raman), and mass spectrometric (e.g. time of flight (TOF) and tandem mass spectrometry (MS/MS)) techniques [1–3]. Each technique has its benefits and disadvantages. Spectroscopic techniques can often not differentiate between similar compounds.However, the technique is non- or less invasive than chromatographic approaches, and requires little sample preparation. Whereas, chromatographic techniques are invasive and require an extensive sample preparation but have an increased resolving power and can easily be coupled to a mass spectrometer to obtain high sensitivities. The relatively small amount of sample available, the low concentration of the dyes, and the even lower concentrations of degradation products contribute to an intricate decision when assessing the most suitable analysis strategy for photodegradation samples. To be successful, it is important to know what type of photodegradation products are generally being formed.

In this literature study, different approaches for the analysis of photodegradation products and their degradation mechanisms will be discussed for one group of natural dyes, flavonoids, and one group of synthetic dyes, triarylmethane dyes. Therefore, as part of the study, the principles of photodegradation and accelerated light ageing will be explained. Next, the general photodegradation mechanism of flavonoids and triarylmethane dyes will be reviewed to highlight the expected products during photodegradation. Finally, the different analytical techniques will be assessed and compared for their feasibility in photodegradation studies. Advantages, drawbacks, and the experimental setup, will be discussed and summarised in the conclusion.

(5)

5

2. Natural and synthetic dyes

A dye is an organic compound used to give color to various substrates. Dyes can be distinguished between natural and synthetic based on the source of the material. Natural dyes originate from natural sources such as plants and insects, and have been used since antiquity [4,5]. Synthetic dyes were developed in the late 19th_{century and make up for most of the colors}

that we see today. Synthetic dyes are cheaper to produce, brighter, and easy to apply to the fabric. Apart from their origin, dyes can be categorized based on different structural properties (azo, anthraquinone, flavonoids, indigo, phthalocyanine, sulfur, nitro and nitroso dyes) or according to their application method (disperse, acid, basic, direct and vat dyes) [6]. Both structural differences and binding interactions with a substrate or solvent matter for the lightfastness of dyes. The photodegradation of different dye classes are difficult to compare and only few dyes are extensively studied and elucidated. Therefore, this literature study will focus on one synthetic dye class, triarylmethanes and one natural dye class, flavonoids.

2.1 Triarylmethane dyes

Triarylmethanes are one of the earliest classes of synthetic dyes ever produced. Their bright and versatile colors ranging from red-purple to blue-green became widely used near the end of the 19th_{century [7]. The dyes are used in a variety of applications such as, textile dyeing,}

ballpoint and felt tip inks, printing, and as staining agent in bacteriological and histopathological processes. The triarylmethane dyes are defined by their chromophore system of three conjugated aromatic rings bonded to a central carbon atom [6]. Triarylmethane dyes can be grouped in different subclasses [8] (see table 1), according to the nature of the substituents and positions on the aryl groups. The dyes are characterized by their poor light fastness and are extremely fugitive. An example is the use of Crystal Violet by Vincent van Gogh in drawings and letters such as, ‘Montmajour’ (1888 Arles) [9], in which the purple color was degraded to a brown color, and ‘Menu’ (1886 Paris) in which the dye completely degraded making the letter illegible [10]. The triarylmethane dyes used for art and ink purposes are often complex mixtures of homologues compounds, often differing only by the position and substituents of the aromatic rings. The similarity in the compounds complicates the identification of historically known dyes such as methyl violet, a mixture of tetra-, penta-, and hexa- methylated pararosaniline [9] (see figure 1A); and fuchsine, a mixture of pararosaniline, rosaniline (magenta I), magenta II, and new fuchsine (magenta III), see figure 1B. This becomes increasingly more difficult when the homologues are similar to the degradation products.

(6)

6

Figure 1 Chemical structures of the components of A. Methyl violet dyes, and B. Fuchsine dye. Reproduced from [7].

Table 1 Different subclasses according within the triarylmethane dyes [8], a description of their composition, and the commonly known dyes within the subclasses.

Subclasses Nature of the substituents and structural

differences Commonly known dyes within the family Methyl violet

dyes Dimethylamine groups on the p-position of the aryl groups Methyl violet 2B Methyl violet 6B

Crystal violet (Methyl violet 10B)

Fuchsine

dyes Primary or secondary amines on the p-position of the aryl groups Pararosaniline (Basic Fuchsine) Rosaniline (Magenta I) Magenta II

New Fuchsine (Magenta III) Phenol dyes Two or more hydroxyl groups at the p-position of the

aryl groups. Various substituents such as sulfides or halogens on the aromatic ring.

Phenolphthalein Phenol red

Bromocresol green Malachite

green dyes Methyl violet dye with a singular phenyl group instead of an aryl group Malachite green Brilliant green Victoria blue

dyes Methyl violet dye with a Naphthylamine group instead of an aryl group. Various substituents on the secondary amine of the naphthyl group and the amine on the phenyl rings

Victoria Blue B Victoria Blue R

Xanthene

dyes1 Linkage of 2 aryl groups through oxygen forming a _{xanthene core and a carboxyl on the o-position of the}

3th aryl group. Various substituents including halogens, nitrogen dioxide and alkanes.

Eosin B Eosin Y Rhodamine B Fluorescein

1_{Xanthene dyes are not always classified as a triarylmethane dye because of the xanthene core. In this literature}

(7)

7

2.2 Flavonoids

Flavonoids are a group of similar structured natural substances found in plants, fruits, flowers, roots, vegetables, tea and wine [11]. Flavonoids are responsible for most of the natural yellow colorants in paintings, lakes and cultural heritage objects [12]. In addition, flavonoids are associated with multiple health-promoting effects due to their oxidative, anti-inflammatory, anti-mutagenic and anti-carcinogenic properties [13]. Chemically the flavonoids can be subdivided into different groups depending on the degree of unsaturation and oxidation of the C-ring, the carbon on which the B-ring is attached (see figure 2) and the structural features of the C-ring. These subgroups are flavones, flavanols, flavanones, flavanonols, catechins, anthocyanins, chalcones and neoflavonoids [11]. The active color component of yellow natural dyes are mainly flavones, apigenin and luteolin, and flavanols, quercetin and kaempferol [14]. These flavonoids occur in nature as sugar derivatives, mainly glycosides, and are extracted from plants such as weld, young fustic, dyers broom and sawwort [15]. Because of the wide variety of plants available for dye extraction, no single source of yellow became pre-dominant throughout history. Flavonoids are not only used as yellow pigments, but the flavonoid-dyes are also commonly mixed with the blue dye indigo to produce green [16], and the structural similar anthocyanins give a blue, purple, reddish color [17]. Dyeing with flavonoid dyes requires the use of mordants to ensure a reasonable color fastness to washing and light [18]. Despite various research on the analysis of flavonoids and dyes it is still difficult to unequivocally identify the dyes due to the different plant sources, the relative instability of the lakes and the low quantities of organic molecules present in the dyes [16].

(8)

8

3. Photodegradation and artificial light ageing

The color of a dye arises partly through the absorption of light. Depending on the chromophore, photons of different wavelengths are absorbed and others are reflected, resulting in the color that we can observe. The absorption of light promotes the dye molecules towards a more reactive excited state. The pathway through which the molecule loses the excess absorbed energy defines whether photodegradation will occur and in which extent. There are many competing pathways influenced by a considerable amount of different factors for an excited dye to lose its energy. This chapter will describe the basic types of photochemical conversion and address the main parameters affecting dye photodegradation, discoloration and photodegradation studies.

3.1 Photodegradation

Photodegradation is the photochemical alteration of materials based on the absorption of light. A photochemical reaction requires the excitation of an electron from a ground state orbital to an excited state orbital. Most excited molecules will not undergo photochemical reactions and will lose the excess energy through physical processes. This can happen through three fundamental principles; (1) radiationless transitions: after intersystem crossing or internal energy conversion the energy is lost through vibrational relaxation; (2) emission of radiation: loss of energy through photoluminescence; (3) intermolecular energy transfer: the energy is transferred to another molecule which is a type of photosensitizing [19].

Primarily, photodegradation reactions originate from the triplet states which have longer lifetimes compared to the singlet states. The longer a dye remains in the triplet state, the higher the change of photodegradation and thus a higher rate constant for intersystem crossing. The primary processes of photochemical reactions in solution are depicted in figure 3. Note that the reactions are based on in-solution degradation, the influence of the type of solutions and the comparison to a substrate based degradation will be addressed later on in this chapter. Two types of reactions are classified, according to whether the reaction involves radical species. Type I: the primary reaction products are radicals. A photosensitizer reacts with a compound or substrate through electron transfer or hydrogen abstraction to produces free radicals [20]. Type II: reaction does not involve radical species. A sensitizer in excited state transfers its energy to ground-state molecular oxygen producing singlet oxygen (1_O₂_{), an extremely reactive}

oxygen species [20]. Most photodegradation processes occur through photo-redox reactions and are dependent on the interaction with a reductant or oxidant. An important parameter in this process is the presence of free oxygen molecules. In aerobic conditions the degradation will predominantly occur through a type II photo-oxidation with singlet oxygen. Whereas, in anaerobic conditions both reduction and oxidation reactions can take effect depending on the experimental conditions. Other possibilities for both aerobic and anerobic conditions is intramolecular rearrangement through excited-state intramolecular proton transfer (ESIPT) or dye-dye interactions [19,21]. Whether intramolecular rearrangement results in more or less photostable products, depends on the initial dye structure and the solvent properties. Furthermore, change in substituents of a given dye can result in a switch of the predominant mechanism. In particular electron-donating substituents promote oxidative pathways, whereas withdrawing substituents accommodate reductive routes. Often the main photodegradation products can be explained through the combination of photo-reduction and photo-oxidation pathways.

(9)

9

Figure 3 Primary pathways for the loss of excess energy in an excited dye molecule. D = dye molecule, D*=

excited dye molecule.

Photo-reduction pathways

The photoreduction pathways involve radicals and proceed as type I reactions. The formation of radicals requires high energy light intensity to facilitate bond-dissociation. The reduction pathways initiate via hydrogen abstraction by the excited dye from a neighboring reductant (𝑅𝑅), e.g. functional groups on the textile or contaminants, forming H2D▪ and R▪ radicals (scheme 1).

Afterwards, the H2D▪ and R▪ radicals can react as depicted in scheme 2 and 3 promoting further

photodegradation. Photo-reductive pathways are more likely to initiate decolorization through small molecule losses compared to destruction of the chromophore.

𝐻𝐻𝐻𝐻∗_{+ 𝑅𝑅𝐻𝐻 → 𝐻𝐻}

2𝐻𝐻 ▪ + 𝑅𝑅 ▪ (1)

𝐻𝐻2𝐻𝐻▪→ Decomposition (2)

𝑅𝑅▪_{/ 𝑅𝑅𝑅𝑅}

2 ▪ + 𝐻𝐻𝐻𝐻 → Decomposition (3)

If there is free oxygen available, the H2D▪ radical can also react with oxygen enabling type I

photo-oxidations (scheme 4). The dye acts as a sensitizer and is not consumed during the process, but will initiate further degradation. It is also possible for the R▪_{radical to react with}

oxygen, resulting in the formation of peroxy radicals (RO2 ▪) (scheme 5). The peroxy radicals

are highly reactive and will likewise promote further degradation. Peroxy radicals can also be formed through the dissociation of hydrogen peroxide (H2O2) when present in the solution.

Hydrogen peroxide is an important initiator in the dye degradation of wastewater due to the strong oxidation capabilities.

𝐻𝐻2𝐻𝐻▪+ 𝑅𝑅2 → 𝐻𝐻𝐻𝐻 + 𝐻𝐻𝑅𝑅2 ▪ (4)

𝑅𝑅▪_{+ 𝑅𝑅}

2 → 𝑅𝑅𝑅𝑅2 ▪ (5)

Radicals can also be generated through direct bond-dissociation of the dye, scheme 6, where 𝑅𝑅1▪and 𝑅𝑅2▪ represent radicals produced by the dye itself [22]. These can then further react in

similar reactions as illustrated in scheme 2&3. 𝐻𝐻𝐻𝐻∗ ▪_{→ 𝑅𝑅}

(10)

10 Photo-oxidation pathways

When the dye is excited to a triplet state (scheme 7), it can undergo a triplet-triplet annihilation with oxygen, which has a triplet ground state, producing singlet oxygen (scheme 8). The production of singlet oxygen requires a photosensitizer, which can be another absorbing compound or the dye itself. The sensitizer transfers its energy from the lowest excited triplet states to the acceptor. In this case, the molecular oxygen is promoted to an excited state. Most dyes will be degraded by singlet oxygen when there is free oxygen present in the solution (scheme 9) often resulting in destruction of the chromophore [22].

𝐻𝐻 + ℎ𝑣𝑣 → 𝐻𝐻1 ∗ _{→ 𝐻𝐻}3 ∗ ₍₇₎ 𝐻𝐻 3 ∗_{+ 𝑅𝑅}3 2 → 𝐻𝐻 + 𝑅𝑅1 2 (8) 𝑅𝑅 1 2 + 𝐻𝐻 → 𝐻𝐻𝑅𝑅2 → Decomposition (9)

Oxygen may also engage in a type I photo-oxidation [22,23]. The quenching of excited states of dyes by oxygen can induce electron-transfer reactions, resulting in the formation of superoxide (scheme 10) [24]. Subsequent reaction of superoxide with the dye induces further decolorization and decomposition (scheme 11) [25,26]. The formation of superoxide is rare in solution and usually the type II process is preferred, especially at low dye concentrations and low-energy excitation wavelengths [19,27].

𝐻𝐻∗_{+ 𝑅𝑅}

2 → 𝐻𝐻+▪+ 𝑅𝑅2−▪ (10)

O2− ▪+ 𝐻𝐻 → Decomposition (11)

3.2 Oxygen

As seen in the previous paragraphs, oxygen is undoubtedly an important parameter for the photodegradation of dyes. The presence of free oxygen opens up different pathways and distinctly active during photodegradation. The question remains how does changing the amount of oxygen impact the overall loss of color of the dye?

Generally removal of oxygen decreases the photofading of dyes [28–31]. Even though reductive (and potentially oxidative) pathways will still occur, the overall loss of color is significantly reduced. This is mainly because the photo-oxidation through singlet mostly destroys the chromophore with as consequence complete loss of color. Whereas, with reductive pathways the chromophore will often likely stay intact resulting in less pronounced photofading and a small shift in the excitation wavelength [32]. However, this is not always the case and often the literature shows contradicting results. An example is the photodegradation of carminic acid which is a natural anthraquinone dye. The photostability of anthraquinone dyes is poor in the presence of oxygen and singlet oxygen sensitizers are known to increase the photo-oxidation of carminic acid in aqueous solutions [33–35]. A logical conclusion based on these observations would be to preserve cultural heritage objects containing carminic acid in anoxic conditions to decrease the photodegradation. Nonetheless, are anoxic conditions not necessarily better as the possible degradation mechanisms are also dependent on the influence of the solvent, mordant, substrate, and the structure of the dye [36,37]. Studies on the photodegradation of carmine lake monitored with vibrational spectroscopy indicated enhanced degradation of carminic acid under anoxic conditions [37,38]. In contrast, other studies observed a decrease in the photodegradation of carminic acid in carmine lake paint samples

(11)

11 with anoxic conditions [29,39]. Because of the multitude of parameters, it is not possible to draw universal applicable conclusion on the comparison of oxic and anoxic conditions. In spite of the complexity on the influence of oxygen and the contradicting result found in literature, there are agreements on the comparison of medium and high oxygen concentrations. The rate of photo-oxidation increases with the partial oxygen pressure at low oxygen concentrations but flattens out around medium oxygen concentration (atmospheric oxygen, roughly 20%) [33,37]. Increasing towards 100% oxygen content does not further increase the photo-oxidation rate. This rate can be further enhanced through more efficient photosensitizing. It is also important to note that oxygen, e.g. atmospheric oxygen, can also oxidize dyes without the presence of light [40,41]. Resulting in complications when calculating rate constants; for example, because of the degradation of stock solutions [42]. Associating the degradation products solely to the light-induced pathways is rarely achievable.

3.3 Solvent

The solvent and pH play a role in the protonation and deprotonation of redox active groups influencing the photo-redox rates [14,43]. On top of that the relative lifetime of singlet oxygen is increased in aprotic solvents such as dimethyl sulfoxide (DSMO) and acetonitrile (ACN), increasing the chance of photo-oxidation [24,44,45]. But most importantly, solvent interactions influence the photorearrangement and the excited-state proton transfer (ESPT) rate of dye molecules. Most natural dyes, such as anthraquinones and flavonoids, have tautomeric forms when radiated with light. Via ESIPT the energy in the Locally Excited (LE) form can be transferred to form a photo-tautomer (PT). ESIPT is very sensitive to competition and interference from intermolecular ESPT with solvent molecules [46]. Aprotic solvents, such as DSMO and ACN, and non-polar solvent such as alkanes, enable rapid proton-transfer kinetics resulting in an increased ESIPT. Whereas, polar protic solvents such as water and alcohols hampers ESIPT and increase intermolecular ESPT [47]. The hydrogen bonds from polar solvents form a complex with the dyes, increasing the energy barrier between the LE and PT tautomer and thus reducing the ESIPT [48].

ESIPT can have both a positive and a negative influence on the photofading, depending on the structure of the dye. For example, in the case of anthraquinones, an increased ESIPT decreases photodegradation [43]. The excited state of an enolic tautomer relaxes to the keto form in the ground state resulting in less photodegradation (figure 4). The energy is lost through relaxation without a photochemical change [46]. For most flavonols, an increased ESIPT rate can also result in more photodegradation. The tautomeric states of flavonols are prone to rearrange intramolecularly, resulting in penthatomic structures and thus loss of color [49]. In this case changing the solvent would be a tradeoff between favorable ESIPT and unfavorable rearrangements. (The photodegradation mechanism of flavonols will be further discussed in chapter 4). An increased ESPT rate in polar solvents also means that the dye will be more inclined to act as photosensitizer for other photochemical conversion processes; such as type II photo-oxidation [47].

It is also possible for the solvent to react during the photodegradation. Dall’Acqua et al. [50] detected the solvent addition of ethanol to quercetin and Pirok et al. [51] observed reactions between DSMO and the carboxylate group of carminic acid and eosin. Altogether, the exact influence of the solvent on the photodegradation reactions are complex and often difficult.

(12)

12

Figure 4 Photo-tautorism of alizarin with on the left the enolic tautomer (LE state) and on the right the keto tautomer (PT state).

3.4 Substrate

Comparing the photodegradation of dyes in solvents and on a substrates, e.g. a textile, is challenging. The overall possible pathways of photochemical conversion and the physical processes to deplete energy remain similar. However, reactions with the substrate, the addition of mordants (metal ions), contaminants, and the different dyeing techniques must now be considered when elucidating the photodegradation mechanism. These parameters greatly influence the photodegradation kinetics, ESPT rates, and the impact of structural differences between the dyes.

Dyed textiles show significant changes in the rate of fading depending on their surrounding conditions. Atmospheric contaminants, such as oxides, ozone, nitrogen, and sulfur dioxide, are known to both induce photodegradation reactions and degrade the dye in the absence of light. The extent to which atmospheric contaminants react is also dependent on the distribution, amount and surface exposure of the dye [22,52,53]. Another important factor is the humidity. The presence of moisture facilitates diffusion of reactants, solvate dyes molecules, effect tautomeric equilibria, and can participate in the photo-redox reactions [6]. A low relative humidity (RH) is overall beneficial for dye preservation on textiles but can be unfavorable for textile fibers.

Natural fibers

The substrate to which the dye is bound can undergo excitation and participate in photo-redox, hydrogen abstraction or dissociation mechanism [22]. Different types of fiber with the same dye can substantially alter the amount of photo fading. Protein-based textile matrixes such as wool and silk can absorb ultraviolet (UV) radiation, making them vulnerable for photodegradation and sensitization through their amino acids. Wool can photo-oxidize through histidine, methionine, tyrosine, tryptophan, cysteine and cystine and silk via tryptophan, tyrosine and phenylamine [54]. The photo-oxidation forms radicals such as hydroperoxides and phenoxyl radicals which are responsible for the yellowing of the textile and can induce further photodegradation of the textiles and dye molecules [55–57]. In general, protein-based fibers are more liable for hydrogen abstraction and thus type I photo-redox mechanisms[58]. However, irradiation of wool has also shown to generates singlet oxygen, increasing the potential of type II photo-oxidation [59]. The singlet oxygen formed can also react with the substrate instead of the dye, forming radicals promoting type I photo-redox reactions. Cellulose-based substrates such as paper are relatively stable at lower wavelength resulting in fewer photo-oxidation reactions [19,60]. Reactive dyes are therefore more likely to show reductive fading mechanisms [61]. At near-ultraviolet radiation, the substrate endures chemical deterioration and photolysis,

(13)

13 promoting photo-oxidative processes [60]. Overall the influence of solely the fiber is difficult to predict and dependent on the wavelengths of irradiation [62]. Especially considering whether the dyes are covalently bound or though mordant complexation.

3.5 Mordant

Most natural dyes, like anthraquinones and flavonoids based dyes, will not durably adhere to fibers without the use of a mordant. The word mordant originates from the Latin word "mordere", which means "to bite" and it is used to bind the dye to the fiber. The mordant, most often a polyvalent metal due to its electron-accepting properties [63], forms a coordination complex between the functional groups on the substrate, for example, amino acids, and functional groups on the colorant compounds [64]. Binding between the metal and fiber is mostly through the amino acids: aspartic and glutamic acid, which are especially susceptible to complexation because of their low pKa value of 4, allowing them to be deprotonated at a pH of 3.5-5 [65]. The most commonly used mordant is potassium aluminum sulphate (Al2(SO4)3∙K2SO4∙24H2O) better known as alum, but other salts such as tin chloride

(SnCl2∙2H2O), iron sulphate (FeSO4∙7H2O), copper sulphate (CuSO4·5H2O), and potassium

dichromate (K2Cr2O7) are also frequently used [64,66,67]. The temperature, the dyeing

procedure, and mainly the structure of the dyes affect which complexes form and how effective this formation is [65]. Metal chelation with the colorants occurs mostly through hydroxyl and oxygen groups which can easily be deprotonated [68]. A prime example is the phenolic groups present in flavonoid and anthraquinone colorants. The proton of the phenolic groups is easily displaced at lower pH values; thus a stable coordination can take place. This is not the case for aliphatic alcohols in which the oxygen anion is not stabilized by the mesomeric effect typical of phenols [69]. Bidentate ligands, in which two functional groups can coordinate to the same metal, are more regularly formed than monodentate ligands which only have one available site for metal coordination [69].

Upon complexation, the metal becomes integrated into the delocalized system of the chromophore and lowers the overall energy gap between the π and π* electronic states, altering the color of the dye. Additionally, metal complexation negates ESPT via the bindings in a similar fashion as water or by steric hinder. The better the complexation between the dye and the metal (and at as much chelation sites as possible), the more ESPT becomes impossible [70,71]. Various mordants will have a different result on the energy gap and the hampering of ESPT. Thus, resulting in different light stabilities. This can again result in either a negative or positive influence on the photofading, depending on intramolecular rearrangements.

Metal ions present in the pigment or dye either through mordanting or as contaminant will also influence the photodegradation rates when they are not complexed with a dye. Metal ions can quench excited possibly increasing the light stability [22,72] or can complex with other components present in the mixture. Then, the metal complex can act as a photosensitizer for singlet oxygen or other photo-redox reactions [19,22,73]. Furthermore, transitions metals can catalyze the production of free radicals through reactions with the substrate [70].

(14)

14

3.6 Light intensity

Light is indispensable to observe and enjoy art, yet, simultaneously, an influential factor in its destruction. Dyes can be exposed to various wavelengths depending on the source of the light influencing the photostability. In paragraph 2.1, we already noticed that UV radiation (higher energy wavelengths), promotes photo-reduction reaction through bond dissociation reactions and that lower energy excitation wavelengths stimulate type 2 photo-oxidation processes. We also saw in paragraph 2.4, that protein-based fibers absorb in the UV range, resulting in yellowing and photodegradation of the substrate. These difference between high and low energy wavelengths are generally well understood and applicable to most dyes. The influence of smaller differences in wavelengths, such as between UVA (320-400 nm) and UVB (290-320 nm) are more complicated [74,75]. Comparing exposure of only UVA and only UVB radiation showed similar photodegradation for carminic acid [37], whereas, the photodegradation of turmeric on wool was highly increased in UVA, and significantly less affected by UVB radiation [76]. Each chromophore will absorb light at distinct wavelengths. From a preservation perspective, it is important to know which wavelengths increase photofading for specific chromophores [77]. Unfavorable wavelengths can then potentially be filtered to prolong the lifetime of dyes [75]. Apart from the wavelengths, the intensity of the light also affects the photodegradation.

Accelerated light ageing

To understand photodegradation, accelerated ageing of dyes needs to be performed on textiles, paintings, or in solutions. Ideally, the simulation has similar conditions as normal light ageing. However, realistic studies would be very time-consuming. Based on the reciprocity principle, the simulations are accelerated via higher light intensities and potentially higher temperatures. The reciprocity principle states that the photochemical conversion is equivalent to the net exposure (Wm-2_{) or intensity of illuminance (lux) and the exposure time. Thus, 50 lux of}

illuminance for 1 hour is comparable to the damage produced by 1 lux for 50 hours [78,79]. The reciprocity principle is based on full daylight spectrum and is therefore not necessarily true for situations where the energy emitted by the lamp correlates poorly with solar radiation [80]. There are no standard guidelines for photodegradation studies. Consequently, accelerated photodegradation studies in the literature will show a variety of different artificial light sources like xenon, carbon and mercury lamps and different conditions such as RH, temperature, atmosphere, and wavelength filters [19,51,76–78,80–82]. Various standard methods do exist for assessing the rate at which a dye discolors such as the AATTC blue wool standards and the grey-scale standard [83–85]. The tests compare the fading of the sample to a scale of standard reference samples each correlation with a certain amount of photodegradation. Although the tests are frequently used, there are still questions regarding the applicability in different light sources [74,80,86]. Another issue with these test is that the visual interpretation is operator dependent. This may require colorimetric measurements to provide an objective comparison.

(15)

15

4. Photodegradation mechanism and products

In order to analyze the photodegradation of dyes, it is imperative to know what products could potentially form during the photodegradation process. This is even more essential for cultural-heritage objects as the original chromophore and possibly its degradation products may already be degraded. The multitude of external conditions, discussed in chapter 2, make it difficult to compare the photodegradation mechanism of different studies and dyes. In this chapter, the photodegradation mechanism of flavonoids and triarylmethanes dyes will be described and compared. The emphasis lies on possible intermediates, the photodegradation products and trends of the different dyes. This will include a brief overview of the photodegradation mechanism and will not include for current theories on all possible photodegradation mechanisms and their reaction kinetics.

4.1 Photodegradation mechanism triarylmethane dyes

The majority of research on the photodegradation mechanism of triarylmethane dyes have been applied on methyl violet and particularly crystal violet dyes. Their various uses in art, forensic and biological applications contribute to a generally accepted degradation mechanism. In addition, crystal violet is often the choice for photodegradation studies because of the high resemblance to the other dyes in the family. Fuchsine, malachite, eosin, rhodamine and phenol dyes are studied in a lesser extent and their mechanism is often not completely elucidated. The photodegradation of triarylmethane dyes can be categorized through three fundamental pathways.

Loss of substituents

All triarylmethane dyes can lose their substituents via type I photo-redox reactions. For methyl violet, fuchsine and malachite green (MG) dyes this results in N-demethylation [3,82,87,88]. The methyl group is sequentially replaced by hydrogens [89] see figure 5. Demethylation results in absorption at shorter wavelengths (blue shift). Ethyl groups will react in a similar fashion, if present on the dye, as is observed with rhodamine B and brilliant green [90,91]. The demethylation is solvent depended. Weyermann et al. [92] demonstrated that CV demethylated in water but not in ethanol under the same conditions indicating different pathways and kinetics depending on the medium. They hypothesized that the demethylation in water is linked to the formation of hydroxyl radicals whereas with ethanol, the produced radicals should enable photooxidative electron/proton transfer mechanisms. Deamination was also detected for methyl violet dyes presumably after demethylation [9,10] and occurred more frequently during catalyzed photodegradation [93]. N-oxides and N-imides were identified as product of reactions and rearrangements via the amine but the reaction mechanism remains unknown[10].

Eosin, a halogenated triarylmethane, experiences substitution reaction (SN1) resulting in the loss of the bromine (see figure 6) [94]. The products are predominately mono debrominated, but double debrominated species can occur depending on the condition [32]. All bromine substituents can participate in the SN1 reaction resulting in isomers [51]. Dehalogenation degrades the dye, giving rise to a shift in the wavelength but will not completely remove the color. Other triarylmethanes with halogen auxochromes are expected to react similarly [94].

(16)

16

Figure 5 Light assisted demethylation of crystal violet. Reproduced from [88].

Figure 6 Possible reaction mechanism for the loss of bromine via photo-oxidation. Reproduced from [50].

Photo-oxidative cleavage

Photo-oxidative cleavage of the central C-phenyl bonds is the main pathway for destruction of the chromophore, resulting in benzophenone and phenol formation, see figure 7 [89]. The photo-oxidation proceeds through the attack of singlet oxygen molecules [95], which are known to form during photolysis of triarylmethane dyes in solution and on substrate [30,95]. Another possibility is a ring opening reaction with hydroxy radicals formed by singlet oxygen (in water) resulting in similar degradation products [30,94]. The presence of singlet oxygen sensitizers, e.g. methylene blue and titanium dioxide, accelerate the photo-oxidative cleavage and the process is diminished when singlet oxygen quenchers are present. This confirms that singlet oxygen is involved in the photo-oxidation pathway [92].

Figure 7 The mechanism of the degradation of methyl violet through the attack of singlet oxygen. Reproduced from [88].

A particularly interesting benzophenone in the degradation of methyl violet, fuchsine and malachite green dyes is Michlers ketone (MK) and its derivatives. MK is used in the synthesis and can be present in the beginning of the dye/ink mixture [8] but is also believed to be a degradation product of methyl violet dyes [6,9]. In addition, MK accelerates the photodegradation of triaryl methane dyes through various ways of sensitizing, energy transfers from the triplet states and the formation of radical species which may further evolve into degradation products [10]. The exact involvement and which of the process occur, or even all

(17)

17 combined, is yet to be determined. However, it is known that MK is consumed during the process and hence, does not functioning as a catalyst.

Xanthene dyes are stabilized by the xanthene core. Therefore, xanthene dyes such as rhodamine and eosin show less photo-oxidative cleavage reactions. Xanthene dyes are more presumed to photodegrade via ring opening reaction involving hydrogen peroxide and hydroxyl radicals. The pathway for the cleavage of the chromophoric moiety at the center of the molecule is still unclear and not fully understood [96].

Leucoside form

Most triarylmethane dyes can be converted into a non-ionic colorless form, leucocarbinol. The transformation is caused by heat, pH or light and often reversible such as with pH indicators. Photoreduction of the excited dye can also cause a permanent colorless leuco form, see figure 8 [92,95,97]. Photoreduction of the excited dye cation is achieved through the addition of an electron or by photochemical hydrogenation of the dye. The leucoside is colorless as the conjugated system is disrupted by the formed Sp3 carbon. It is important to note that the leucoside is still (or partially) methylated.

Figure 8 Example of the conversion of malachite green between its colorless leuco forms. Reproduced from [96].

All the above mentioned photodegradation mechanisms may occur competitively with each other. Exact contributions of which degradation pathway and the kinetics of the degradation are therefore extremely complex. These fundamental mechanisms do not account for all the current theories on possible alternative degradation pathways and all of the observed products, but only focusses on the mainly observed products.

4.2 Photodegradation mechanism of flavonoid dyes

Flavonoid compounds are known to have poor lightfastness, making them susceptible to photochemical processes like photo-oxidation [98]. The photodegradation mechanisms of flavonoids, and particularly which intermediates form and how this influenced by the experimental set up, are still ambiguous. In consequence, most flavonoid degradation studies focus on comparing structural properties of different flavonoids and their photostability without a distinct connection to a mechanism. The photostability varies depending on the structure and most importantly the C3 position of the flavonoid [99,100]. Multiple studies highlighted the presence of a hydroxy group on the 3 position as the determining factor in flavonoid photo-reactivity [41,49,101,102]. Substitution of the 3-hydroxy group with a sugar moiety stabilizes the flavonoid towards light [101,102]. Other hydroxy groups in the 3’ and 4’ positions were also found to marginally influence the photostability [49], whereas, hydroxy groups in the

(18)

18 positions 5 and 7 were observed to not play a significant role in the photodegradation mechanism [49,50]. The photodegradation products vary from low molecular weight compounds such as dihydroxy benzoic acids [41,98,99], dimers and oligomers [14,41], small molecule loss [103] and rearrangements products [41,47,104]. Establishing a link between the structural properties, products, and the photodegradation pathways is challenging. Nonetheless, there are two hypothesized pathways; photo-oxidation through depside formation and photorearrangement, which can explain most of the degradation products.

Photo-oxidation

Oxygen is important factor in the photostability of flavonoid compounds [42,105]. The main photodegradation products in oxic conditions are hydroxy benzoic acids [41,98,99]. The amount of oxidative reaction products increases with exposure time to oxygen indicating type II photodegradation reactions. The photodegradation mechanism is hypothesized to proceed via the formation of a depside though the photo-oxidation of the C2-C3 linkage, followed by the breakage of the C3-C4 bond [98,99]. Although, it is evidential that the reaction occurs through oxidation of the C2-C3 link, the depside formation and the potential catalyst for the reaction remain uncertain. Colombini et al. [99] suggested two pathways, catalyzed via a metallic ion (figure 9 a) or through type II photo-oxidation (figure 9 b). The hydroxybenzoic acids are an important molecular marker for the studies of degraded artwork. As the position of the hydroxyl group depends on the nature of the characteristic flavonoid chromophore of the dye these contain information on the original dye source. Molecular markers have been observed for morin, quercetin and kaempferol, see figure 10.

Figure 9 Proposed degradation mechanism of morin a. catalyzed by a. mordant and b. through light induced photodegradation [99].

(19)

19 Photorearrangement

Photo-rearrangement can take place from a singlet excited photo tautomer (1_{PT). Protic polar}

solvents, such as alcohols, decrease rearrangements from the 1_{PT and inhibit the}

photorearrangement, subsequently it will enhance photo-oxidation from the 1_{PT. Aprotic polar}

solvents have the opposite effect and do not interfere with the photorearrangement of 1_{PT. The}

flavonoids can still undergo oxygenation afterwards because of a rapid 1_{PT to}3_{PT conversion,}

but will also show penthatomic structures (figure 11). The photorearrangement is extensively studied for various flavonoids, with different substituents, and in multiple solvents [41,47,106]. The rearrangements vary in structure and the rate is heavily dependent on the used conditions. Because of the different conditions and simplistic studies it is difficult to apply this knowledge to the photodegradation of cultural heritage objects. However, it is obviously a prominent factor in understanding flavonoid degradation.

Figure 11 Photorearrangement of flavonol in ACN/DCM to a penthatomic structure [46].

(20)

20

5. Analysis of dyes and photodegradation products

Characterization of dyes and its photodegradation products and processes is challenging because of the relatively small amount of sample available, the low concentration of the original dye content, and the even lower concentrations of degradation products. The analysis is more complicated for cultural heritage objects as artworks may contain other compounds such as fillers, additives, binding media, and other impurities. Ideally, the analysis technique would be rapid and robust to accommodate for the large number of experiments needed to elucidate all the different parameters affecting photodegradation. However, simultaneously, the method needs to be sensitive to measure the low concentrations especially for cultural heritage objects that have limited sample available. This is true to a lesser extend for simplified mock-up photodegradation studies but still relevant. In the following chapter, the different analytical techniques used in photodegradation studies will be reviewed and their limitations will be discussed.

5.1 Liquid chromatography

By far the most applied technique in the analysis of dyes and photodegradation studies is liquid chromatography [1]. Since most dyes are water soluble organic compounds they are particularly amenable to being separated by high-performance liquid chromatography (HPLC), more specifically reversed-phase. Because dyes contain strong chromophores detection is often performed through their absorbance in the ultraviolet and/or visible regions of the electronic spectrum. An important development for the analysis of dyes was the diode array detector (DAD), also known as the photodiode array detectors (PDA). DAD allows for simultaneous detection at multiple wavelengths, crucial for the detection of dyes sample as they may absorb photons at different wavelengths. While HPLC-DAD is often used, it does lack sensitivity for the precise identification of dye components [5]. UV/DAD also suffers from low specificity in the characteristic UV-vis spectral shapes as co- (or partially) eluting peaks cannot be discriminated from each other. Identification can be performed based on the HPLC retention time if standards are available, however, this is often not the case and tends to be expensive [15,107,108]. Also, the UV-Vis absorption of chemical compounds is affected by the mobile phase composition and the optimal absorption wavelength of the dye will likely shift as the molecule structure is altered during the photodegradation process [51]. To unambiguously identify the dyes and photodegradation products mass spectrometry is required.

Mass spectrometry

Mass spectrometry provides additional information about the molecular structure, the mass-to-charge ratio (m/z) of molecular ions, and its fragments [109]. In combination with HPLC, this information offers additional discrimination power for the identification of dyes that cannot be reliably distinguished on only their UV-Vis absorption spectra [5]. MS cannot be used for qualitative and quantitative analysis of complex mixture without additional separation due to adverse effects in the ionization chamber such as discrimination and ionization suppression [110]. In addition, mass spectrometry by itself only has limited isomer identification potential as only the fragmentation pattern can be used to distinguish similar mass products [5]. MS detection, and especially tandem MS detection, offers better limits of detection (LODs), improved selectivity, and allows for structural elucidation of unknown components without the use of standards[5,111]. MS detectors are often used combined with DAD detectors for the

(21)

21 analysis of dyes [1,5,112–115]. The techniques provide simultaneous analysis of the retention time, UV-Vis absorption spectra, and the molecular characteristics. Another benefit of using LC-DAD-MS is that the technique can screen hundreds of extracts and gain partial identification on most of the compounds [116]. Tandem MS can provide additional structural information and allows for the comparison and matching of the fragmentation characteristics of certain structural features, however, this would reduce the sample throughput [114,116]. LC-MS flavonoids

There are many different flavonoid dyes from different sources [116]. Most of these flavonoids are glycosides consisting of a sugar attached to an aglycone flavonoid. HPLC-DAD alone is not adequate to measure dye and plant samples that contain similar flavonoids as the UV-Vis spectra within a class are indistinguishable [116]. Thus, MS is necessary for proper flavonoid identification. Dyes from a plant origin usually contain many derivatives of the same family of compounds and hence isomers with a different retention time but the same mass occurs often. These cannot be identified using only the molecular mass of the compound as this can lead to misinterpretation of compounds from dye sources. Therefore, flavonoid analysis is often performed with LC-MS, LC-DAD-MS, or LC-MS/MS.

The mass spectral characteristics of flavonoids and particularly that of the electrospray ionization (ESI-MS) coupled with LC have been extensively studied [13,16,114,117]. These studies came to the conclusion that optimally flavonoids should be measured in both positive and negative ion monitoring. Electrospray ionization mass spectrometry provides a highly sensitive technique for the analysis of flavonoids in solution and is therefore preferential for the analysis of natural yellow dye extracts from textiles and lakes. Soft atmospheric pressure ionization techniques have also been applied for the analysis of the glycan sequence. The patterns observed with soft ionization techniques allows one to easily determine the molecular weight of the compound. This is useful for the identification of the glycosidic and oligoglycan moieties and therefore to determine the glycan sequence [113]. The type of Mass detectors used combined with LC flavonoid studies are depicted in figure 12. TOF is the most popular due to its performance characteristics such as high speed, good resolving power, and relatively low cost compared to HRSM systems. Q-TOF allows for the addition of tandem MS and high resolution data making the technique more suited for structural elucidation, especially for unknown structures. An overview of the mass analyzers is presented in appendix 1. The LC separation mechanism used for flavonoid analysis is for 98% RP-LC and 2% HILIC [13]. As a simple C18 column can perform the separation of flavonoids dyes without significant issues most photodegradation studies tend to not optimize the separation.

(22)

22

Figure 13: Overview of the application of different MS instruments used for LC-MS analysis of flavonoids during 2009-2015 reproduced from [13]. LIT: linear ion trap, IT: ion trap.

While optimized methodology with different mass analyzers are readily available for flavonoid analysis, few studies focus on both flavonoids and their degradation products. The MS techniques and LODs of studies containing flavonoid and degradation products reported in the literature are summarized in table 2. These studies tend to use a targeted approach through selected ion monitoring (SIM) or multiple reaction monitoring (MRM) modes to increase the sensitivity. In SIM mode, only preselected m/z values are detected in the analysis. In MRM the initial target ion gets fragmentated and subsequent selection of the daughter fragment ions are used for the quantitative purposes, whilst ignoring all other ions in the mass spectrometry. Degano et al. [118] compared MRM and SIM mode for the analysis of flavonoids and its known degradation products. They observed LODs in the same order of magnitude for both MRM and SIM mode. MRM was deemed preferential due to its capability to distinguish isomers. Surowiec et al. [103] compared both DAD and MS with positive and negative ionization in SIM mode for the detection of flavonoids, anthraquinones and hydroxybenzoic acids. Comparison between positive and negative ionizations modes showed that negative ionization mode produced better signal to noise ratios than positive mode especially for the hydroxybenzoic acids. They determined LODs in the same order of magnitude for both DAD and MS in sim mode, the table containing the LODs and a list of the studied compounds is added in appendix 2. The LODs were lower with the MS method but the repeatability was also worse with a higher relative standard deviation between measurements.

You can measure flavonoids and their known degradation products satisfactory with a targeted MS approach in negative ionization mode. However, the disadvantage of these targeted approaches is that non selected ions are neglected and thus unknown products can remain undetected. The alternative would be to use a non-targeted approach, in which the full spectrum is scanned and identification is based on the exact mass and MS/MS fragmentation patterns potentially combined with other analytical techniques such as NMR. Unfortunately, the non-targeted approach is time consuming and less sensitive than the non-targeted approaches.

(23)

23

Table 2 The mass spectrometric techniques and modes applied in flavonoid degradation studies.

Technique and

ionization source Ionization and scanning mode Mass analyzers Limits of detection Source

ESI-MS Negative mode Q 5 ng/ml [119]

APCI-MS Negative mode Q 20 ng/ml [119]

ESI-MS Positive mode SIM Q-IT 10-150 ng/ml [120]

ESI-MS Negative mode SIM Q-IT 30-90 ng/ml [121]

ESI-MS Negative mode SIM QqQ 0.5-12 ng/ml [103]

ESI-MS Negative mode MRM Q-LIT 0.4-20 ng/ml [118] LC-MS Triarylmethane dyes.

Various mass analyzers such as quadrupole [122], triple quadrupole [111,123,124], ion trap [9], and time of flight [111,125,126] have been applied combined with LC for the analysis of triarylmethane colorants. Based on the theoretic degradation pathways of triarylmethane dyes it was suggested that isomers would form during the photodegradation, LC-MS allowed for the detection of these isomers [6]. Favaro et al. [10] demonstrated that the degradation products of CV inks or paints can easily be recognized if analyzed with LC-MS techniques with medium-high resolution and good mass accuracy, coupled to DAD detection. MS detection combined with LC also allowed for the tentative identification of N-imido oxides and hydroxylamine derivatives which had not been reported before. Nonetheless, it was still not possible to identify all absorbing compounds not related to CV derivatives from on paper degraded samples due to a lack of standards and knowledge of degradation compounds in old paper due to a lack of standard and knowledge of degradation products from the aged paper being limited. Degano et al.[127] used both positive and negative HPLC-ESI-Q-TOF for the characterization of methyl blue, fuchsine, and methyl violet dyes and its degradation products. Homologous compounds were separated and discriminated based on the fragmentation patterns and specific ions. Using a collision-induced dissociation cell at 50.0 V allowed for different isomers to be distinguished through the interpretation of the fragmentation patterns. The LC separation combined with the strength of the MS detector and the elucidation of the fragmentation patterns makes the technique very suited for complex samples.

Disadvantages of liquid chromatography

There are also distinct disadvantages in using liquid chromatography. Liquid chromatography requires the extraction of solid materials. Firstly, it is micro-destructive, for historical artifacts analysis this requires milligrams of yarn to be removed from the investigated [128]. Secondly, in photodegradation studies it is important that the dyestuff is not compromised by the solvent, which is especially important during the extraction. Extraction often involves strongly acidic conditions such as mixtures of hydrochloric acid, methanol, and water, leading to hydrolysis of glycosidic moieties, esterification, and decarboxylation reactions [129–131]. The stability of the dyes in the solvent and the effect of the extraction solvent should be considered during solvent selections. Particularly the loss of glucoside moieties, often found in natural dyes from plant based extracts, can easily cause misinterpretation in the photostability of a dye as the aglycone variants are often more light sensitive than the glycosidic dyes [101,102,132]. Extraction of textiles can also result in potential impurities further complication detection and the extraction could show bias to specific groups of the analytes, and thus, misrepresent the relative intensity of the dyes and products present in the sample [133].

(24)

24 All things considered, LC-MS can be utilized as a stand-alone technique for the identification of dyes. Preferably the technique is combined with a DAD detector for a more comprehensive analysis. If the LC application requires an extraction, caution is advised to ensure that the sample is unaffected and that the glycosidic content remains unaltered, especially with natural dyes. Both LC-MS and LC-MS/MS proved to be effective for measuring degradation products, however, unequivocal identification still depends on the availability of reference standards or reference libraries, containing mass spectra, fragmentation patterns, and UV-Vis absorbance spectra [109].

5.2 Direct Mass Spectrometry

The last few years saw a rise in popularity of ambient mass spectrometric techniques such as direct analysis in real-time time-of-flight mass spectrometry (DART-MS) [96,127,134,135], laser desorption ionization mass spectrometry (LDI-MS) [136], liquid micro-junction surface sampling probe mass spectrometry (LMJ-SS-MS) [137], matrix-assisted laser desorption ionization (MALDI) [136,138], direct infusion electrospray ionization (DI-ESI) [96], and surface acoustic wave nebulization (SAWN-MS) [139] as alternative techniques for the analysis of dyes in textiles and paper. These direct mass spectrometric methods are minimally invasive as they require small sample sizes and can easily be applied without or with limited sample treatment. These techniques have been utilized in the classification of the type of dyes present in unknown samples via the detection of marker molecules, for instance, the identification of anthraquinones (alizarin and purpurin) by DART-MS [134], flavonoids and indigotins by DART-MS [140], and triarylmethane dyes by DART-MS, DESI-MS and DSA-MS [96,141,142]. Compared to regular LC-MS measurements, direct MS analysis requires a shorter measurement time and does not suffer from any of the potential solubility issues or solvent interactions and void volume elution problems. This comes at the cost of a reduced separation and increased ion suppression effects in complex matrices. In addition, direct MS techniques operate without any separation between isomeric species. Identifying and assigning structures of isomers is therefore often not possible because only the fragmentation pattern is available [5].

Which dyes can be measured with Direct MS techniques depends on the type of ionization used and varies depending on the dye and dye classes. Nguyen et al. [141] measured the colorants present in writing ink with DSA-MS, LC-MS, and GC-MS. They successfully used DSA for the identification of the primary colorants but also concluded that, apart from additional leuco forms, the technique offered little additional information compared to LC-MS. They also noted that the mounting and alignment of the sample was tedious with the DSA-MS analysis negating the benefit of having no sample preparation compared to LC. Drury et al. [142] compared the monoisotopic masses of common colorants with peaks obtained from DART-MS and DSA-MS for the detection and identification of colorants, mostly triarylmethane dyes, in writing inks. Many colorants displayed base peaks in both techniques. However, some only showed base peaks for one of the two methods, see figure 14, and had lower relative intensities in the other method. Armitage et al. [127] observed similar ionization dependencies in the analysis of Peruvian dyes with DART-MS. This method was not capable of detecting carminic acid reliably and only detected trace amounts of the characteristic compounds of cochineal dyes present in the sample. In addition, it was not possible to differentiate between isomeric structures such as alizarin and xantopurpurin with DART-MS. They highlighted the importance of further developing new approaches like ambient ionization mass spectrometry technique for the rapid

(25)

25 classification of dyes but also emphasized the need for established HPLC-MS methods that can provide a comprehensive analysis and characterization.

Figure 13 Colorants identified in ink samples by DART-MS and DSA-MS. The overlapping zone contains the colorants which had base peaks for both techniques; individual zones were base peaks only for that specific technique with lower relative peak intensities for the other technique. Reproduced from [142].

Few studies performed direct MS techniques for the analysis of photodegradation products. Astefanei et al. [139] used the novel and ultra-rapid SAWN-MS for the detection of the main chromophores of triarylmethane and azo dyes and to provide additional information on the degradation products. Through the exact mass (less than 2 ppm accuracy) they identified mono-, bi-mono-, tri-mono-, tetra-mono-, penta- and hexa-mono-, N-methyl-pararosaniline in textiles dyed with crystal violet. The production of ions and the ionization mechanism with SAWN is different from conventional ESI. Still, the fragmentation obtained from Hexa-N-methyl-pararosaniline showed similar fragmentation and product-ions as observed with HPLC-ESI-Q-TOF [7]. SAWN can determine marker molecules with a limited sample size and can be used to track demethylation in simple samples containing triarylmethane dyes. However, this requires further testing on light aged objects for photodegradation studies. Alvarez et al. [96] applied DART-MS and DI-ESI-DART-MS to characterize the degradation pathways of eosin in oil media. They used UVA light to promote debromination via type I pathways and to reduce the structural breakdown of the chromophore. Both DART-MS and DI-ESI-MS were able to follow the debromination pathway during photodegradation. Nonetheless, the techniques were not capable of identifying isomers and showed little of the expected intermediates. In addition, they underline that another pathway is existent that causes the destruction of the chromophoric moiety at the center of the molecule to explain the complete loss of color of eosin. No intermediates or indications of the nature of these photodegradation products were observed with the direct MS techniques.

Another promising application of ambient MS techniques lies in the ability to measure samples in their native states. This allows for the detection of the mordant complexation and can provide information on the effect of the mordant on the photodegradation [96,143]. Furthermore, Armitage et al. demonstrated that DART-MS could be used to measure the glycosides of the main coloring compounds. This allows for better identification and differentiation between colorants from different plant species without extraction or MS/MS.

Overall, direct MS techniques offer advantages when identifying chromophores in the absence of reference standards or as an initial scan to determine the approach for other analytical

(26)

26 techniques. They are particularly interesting for valuable objects with limited sample available. However, the lack of identification and assignment of isomeric structures and difficulties associating with complex samples require the combination of HPLC-MS methods for a comprehensive analysis.

5.3 Gas Chromatography

Considering that most colorants are generally non-volatile, gas chromatographic techniques are rarely exploited for the analysis of natural and synthetic dyes. Due to the high molecular mass and polarity of the target compounds, GC often requires an extra time consuming derivatization step [15]. However, there are distinct advantages for using GC combined with MS. GC-MS provides a good reproducibility and repeatability of the analysis and has the possibility of simultaneous detection of molecules into complex matrices with a single analysis [144]. Another advantage is the availability of extensive libraries for the recognition of unknown compounds such as degradation products, for example, the National Institute for Standards and Technology database (NIST) [145].

Despite the need for derivatization GC has been applied for dye studies for the analysis of anthraquinones [146,147], flavonoids[15,144,148], and indigotins [148]. The studies showed that dyes can be separated and detected with GC-MS and that it was able to detect some minor degradation products (marker compounds) [99]. N,O-Bis-(trimethylsilyl)-trifluoroacetamide (BSTFA) [15,144], trimethylchlorosilane (TMCS) [144] and m-(trifluoromethyl) phenyl trimethylammonium hydroxide (TMTFTH) [145] were successfully used as derivatization agent for dyes. The downside of these derivatization agents is the formation of multiple derivatives for individual compounds, reducing the overall sensitivity of the analysis. The performance of the derivatization agents varied depending on the dye structures studied. Due to the variable amount of reactive functional groups and specifically the hydroxyl groups in natural dyes derivatization required a long time. The reproducibility of the derivatization was relatively poor ranging from 15-25% due to the low stability of most dye derivatives and the lack of suitable internal standards for the derivatization step [15]. Derivatization also breaks the glycosides bonds in natural dyes samples, hampering the use of GC-MS for natural dye identification.

GC-MS has also be used in the analysis of ink degradation in forensic applications [141,149]. The beneficial qualities of GC in forensic ink analysis are mostly due to the identification of ink vehicles and the applicability of ink samples that are not completely dried [141]. However, the studies do show that crystal violet and potential degradation products such as Michlers ketone and diphenylamine can be measured with GC-MS, but also that it is the least informative technique compared with DSA and LC-MS. Nonetheless, they emphasize the importance of multiple techniques for the full spectrum analysis of ink degradation studies. GC-EI-MS analysis of crystal violet photofading in water under solar radiation by Li et al. [93] displayed the same conclusion. They managed to measure fifteen degradation products or intermediates with GC-MS compared to the 49 measured with LC-TOF-MS. Although at first glance this seems very poor the GC information is highly informative, containing mostly low molecular weight products for which LC-MS has a low sensitivity. The use of GC-MS proved that CV degrades further into smaller molecules, alcohols, esters, amines and, benzoic acid (m/z 72-150), presumably through the degradation of Michlers ketone. The smaller molecules contain little information for the identification of dyes in cultural heritage objects and is often

(27)

27 overlooked in photodegradation studies but are valuable for the full photodegradation characteristics of dyes.

To avoid the extra sample pre-treatment, pyrolysis (Py) coupled GC has been practiced for the identification of dye markers. Py-GC-MS can highlight the breaking of the molecular structure. The pyrolysis products of triarylmethane dyes are categorized in databases and extensively studied [150–152]. Even with a database, unambiguous identification is often not possible because of overlapping profiles, non-specific markers, or too low concentrations compared to the binding medium [153]. Triarylmethane dyes are difficult to fragment with pyrolysis and the intact molecules co-elute with their demethylated molecules; thus, Py-GC is not suitable for the identification of photodegradation products.

Altogether, GC applications for dye analysis are limited by the non-volatility of colorants and the need for derivatization but still applicable for photodegradation studies. Specifically for the identification of unknown degradation products and minor components.

5.4 Spectroscopic techniques

An ideal technique for dye analysis in valuable historical objects would be non-invasive and capable of measuring in-situ without affecting the artifact. Therefore, various spectroscopic techniques such as fourier-transform infrared spectroscopy (FT-IR) [154–156], attenuated total reflectance FTIR spectroscopy (ATR-FTIR), Raman [156–159], fluorescence spectroscopy [4,160], and fiber optics reflectance spectroscopy (FORS) [9,161] have been applied for dye analysis. Unfortunately, these techniques can often not provide precise identification of dye chromophores, the structural information is limited and the techniques lack sensitivity. Furthermore, spectroscopic techniques cannot resolve dyes in complex mixtures and the signal is often distorted by matrix components, such as fillers, additives, varnishes, and textile products [162]. Consequently, spectroscopic techniques are not used individually but often as a supplementary technique in photodegradation studies. For example, as initial dye classification [5], to obtain information about the mordant [163], to scan for function groups [164] or to asses dye aggregation [9]. Recent developments in minimally invasive in-situ Raman and surface-enhanced Raman spectroscopy (SERS) methods showed promising applications for photodegradation studies.

Raman

Raman detects the vibrational transitions emerging from the changes in polarizability through irradiation with a high-intensity laser. Raman requires only a small amount of sample and has no sample preparation. Jurasekova et al. [14] researched the effect of pH on the chemical modification of quercetin with FT-Raman spectroscopy. With a simplistic in-situ FT-Raman measurement they were able to distinguish between rearrangement products of quercetins under different conditions. Gorshkova et al. [89] investigated the aging dynamics of triarylmethane dyes in writing inks using Raman spectroscopy. Using Raman, they were able to determine the aging dynamics, interpret and identify the demethylation process of CV, and observe the point at which photo-oxidation occurred during the accelerated light aging. They optimized the laser power, exposure time, and the number of scans and observed no laser-induced degradation during the measurements. The peak intensities in the Raman spectra are dependent on the concentration of the colorants in the focal spot. This is difficult to control during the measurements. Therefore, the ratio of the intensities of the peaks was compared instead of