The difficulty of fully removing green tattoo pigments remains a central challenge in tattoo treatments, along with the potential harm of generated fragments and particles. Although advances from nanosecond to picosecond laser treatments have shown promising improvements, complete removal often still requires multiple sessions. Moving on to femtosecond lasers, which could enable more efficient pigment breakdown, could result in faster and more effective decolorization. However, simultaneously, the production of harmful molecules or the aggregation of particles with aspect ratios close to dangerous ones needs to be monitored. This study was conducted in two phases. First, the discoloration efficiency was evaluated to determine how it compares to currently adopted treatments and, as a consequence, whether higher or lower demanding conditions are necessary. Afterward, the production of potentially harmful fragments was monitored as were the morphology of the produced particles, their aspect ratio, and their abundance. The investigation has been carried out both on green concentrate ink and on the pigment that it contains, i.e., PG7. Furthermore, investigations were carried out at different peak powers with fixed durations as well as at varying durations with a fixed peak power. Comparative studies were also performed on the basis of prior research in our laboratory with other laser types. The analyses are carried out via UV‒Vis spectroscopy, gas‒chromatography mass spectrometry, and associated PCA of the fragments, as well as SEM imaging.

UV‒Vis spectra were taken for all treated samples and for the reference untreated GC and PG7 samples as shown in the four panels of Fig. 1. The spectra are grouped by increasing the laser power at a constant exposure time (Fig. 1A, B) and by increasing the exposure time at a constant power (Fig. 1C, D). The spectra of the GC ink and PG7 before treatment display the typical Soret and Q bands of chlorinated copper phthalocyanines in the ranges of 300–450 nm and 550–750 nm, respectively. The overall intensity of the bands varies with laser irradiation, but the phthalocyanine features, i.e., the B and Q bands, are largely preserved. This implies that the treatments either provoked the fragmentation of the macrocycle into smaller moieties that do not absorb in the UV‒Vis region or affected the aggregation state of the phthalocyanine agglomerates but did not induce partial degradation, resulting in an alteration in the Q/B band ratio.

UV‒Vis spectra of A the GC dispersions irradiated for 25 min at different powers, and B the PG7 dispersion irradiated for 25 min at different powers, C the GC dispersions treated with 3 W of power for various durations and D the PG7 dispersion irradiated with 3 W of power for various durations

From these analyses, two distinguishing features characterize the UV‒Vis spectra of both reference and femtosecond-treated samples. Notably, the spectrum intensity of PG7 is lower than that of the GC ink despite PG7 being a pure pigment, whereas the GC ink contains PG7 diluted with additives and binders. This discrepancy can be attributed to two factors: commercial PG7 although considered pure, may contain additives at unknown concentrations that could exceed those found in the ink formulation. Additionally, the greater hydrophobicity of PG7 than that of the GC ink makes its dispersion more susceptible to particle deposition, effectively diminishing its concentration in the solution.

Furthermore, a second important feature is that the spectral intensity varies linearly with the irradiation power when the duration is held constant at 25 min. In contrast, when the power is fixed at 3 W, the intensity tends to fluctuate as the exposure time increases.

The decrease in the spectral intensity as a function of irradiation power is straightforward since it can be correlated with macrocycle fragmentation into smaller moieties. From a purely mechanistic point of view, this type of correlation suggests a nonlinear optical (NLO) response (as also indicated by the pinkish color around the focus in Fig. SI1 of the Supplementary Information) under the condition of Saturable Absorption (SA) (Li and Li 2012).

To evaluate the discoloration efficiency and compare the Ti:sapphire femtosecond laser with the Nd:YAG nano-/pico- and Ruby nanolasers, samples of GC ink were treated with each of the three laser types.

Figure 2 shows the absorption spectra of samples treated with the Ti:sapphire femtosecond laser (800 nm) and compares them to the spectra of the samples treated with the Nd:YAG nano-/pico- (532 nm) and Ruby (694 nm) nanolasers obtained from previous experiments (Cecchetti et al. 2022c). The experiments were all conducted under the same conditions, using samples of the same ink and concentration, and subjected to treatments with the same total irradiation energy, after control of the composition as performed by Bauer et al. (Bauer et al. 2019).

UV–Vis spectra of the GC ink dispersion upon laser treatment: dark violet solid line = Ti:Sapphire femtosecond laser, red solid line = nanosecond ruby laser, light violet solid line = nanosecond Nd:YAG, light violet dashed line = nanosecond Nd:YAG with array, light blue solid line = picosecond Nd:YAG, light blue dashed line = picosecond Nd:YAG with array. In the inset, the same set of spectra is plotted along with the nontreated sample, reported with a green solid line

These comparisons indicate that the intensity of the spectrum of the GC ink treated with the femtosecond laser is comparable to that of the sample treated with the ruby laser despite high peak power and narrow time structure of the femtosecond laser pulse. To consider the discoloration of GC dispersions treated with a femtosecond laser compared with those treated with Ruby nano-, Nd:YAG nano-, and picosecond lasers, some considerations regarding the decomposition mechanism of phthalocyanine and the effective volume irradiated are necessary.

Since the spot sizes of the laser beam and the vertical dispersion (albeit Gaussian) are different for the three types of lasers, being 2 cm for the Nd:YAG (both nano and pico) and Ruby beams and 1 cm for the Ti:Sapphire beam, the total volume of the GC of interest for the primary beam is also different, with a VN:VR:VTi ratio of 37.5 mm3:37.5 mm3:10.5 mm3, where VN, VR, and VTi are the volumes affected by the Nd:YAG, Ruby and Ti:Sapphire beams, respectively (in conic approximation). Therefore, even when all phthalocyanine molecules in the focus of the Ti:sapphire beam are decomposed, their total amount remains 3.5 lower than that of the Ruby laser.

Additional considerations can be made regarding the overall mechanisms of decomposition. Typically, when immersed in a solvent, chromophores such as phthalocyanines trigger two additional types of phenomena leading to decomposition: cavitation and photothermally induced bond breakage. Although present in all cases, the extent and the efficacy of the phenomena are determined by the wavelength of the laser beam and the absorption spectrum. The lasing of Nd:YAG is carried out at 532 nm, which corresponds to an absorption minimum of the phthalocyanine spectrum as it is the region at 800 nm, where Ti:sapphire lases. The latter, however, is in the infrared region, whose absorption excites vibrational states, thus largely contributing to photothermal decomposition. Notably, additional mechanisms of fast and efficient photothermal conversion are active in solids such as nanoparticles through exciton–exciton annihilation, which results in a high density of excited molecules (Hosokawa et al. 2000). The wavelength of the ruby laser at 694 nm, instead, is the closest to the maximum of the Q band at 646 nm and is more likely to trigger direct decomposition upon electronic excitation.

GC‒MS analysis revealed twelve fragments common to nearly all samples following femtosecond laser treatment with GC ink and PG7, all of which were identified as harmful (Bauer et al. 2022, 2021, 2020; Cecchetti et al. 2022), as illustrated in Table 2, along with their respective acronyms. Trichlorobenzonitrile (TCBN) was not detected in the PG7 samples.

Overall, these compounds can be divided into two different groups: those related to the decomposition of chlorinated phthalocyanine and those stemming from the active chlorination of naphthalene impurities.

The first group comprises chlorinated benzodinitriles, benzonitriles, phthalimides, benzaldehydes, or benzenes. However, some of the fragments were already present in the native GC ink before laser treatment. In particular, pentachlorobenzene and tetrachlorophthalimide were detected in the H2O/ethyl acetate and acetone extracts, whereas pentachlorobenzonitrile was detected in the chloroform extract (Bauer et al. 2022).

The second category is represented by chlorinated naphthalenes, such as 1,4-dichloronaphthalene, 2,7-dichloronaphthalene, 2,3,6-trichloronaphthalene, and 1,3,7-trichloronaphthalene. These compounds likely do not result from phthalocyanine decomposition. Instead, they appear to form through chlorination during laser treatment of preexisting naphthalene or methylnaphthalene impurities present in the original ink (Bauer et al. 2022). This hypothesis is supported by the common use of naphthalenesulfonic acid and its derivatives as surfactants or grinding additives in pigment preparations (Herbst and Hunger 2004).

Interestingly, the types of fragments detected upon femtosecond laser treatment are quite similar to those detected after irradiation with Nd:YAG nanosecond and picosecond lasers (532 nm) at the same total irradiation energy, especially chlorinated naphthalene, tetrachlorobenzodinitrile, dichloro- or trichlorobenzonitriles and tetrachloroisoindolines (Cecchetti et al. 2022). This correspondence strongly supports similarities in the fragmentation mechanism, independent of the excitation wavelength or pulse duration. Instead, chlorinated benzaldehyde is detected solely upon femtosecond laser treatment; however, it might be a derivative of 4-methylbenzaldehyde, which is also present in the native ink composition, rather than a product of phthalocyanine decomposition.

A semiquantitative analysis was performed by plotting the normalized peak areas of different fragments as a function of irradiation time at a fixed laser power (3 W) and as a function of laser power at a fixed irradiation time (25 min). These plots were obtained for GC ink, focusing on six selected fragments: trichlorobenzonitrile (TCBN), tetrachlorobenzonitrile (TeCBN), tetrachlorobenzodinitrile (TeCBDN), pentachlorobenzonitrile (PCBN), 2,7-dichloronaphthalene (2,7-DCN), and 2,3,6-trichloronaphthalene (2,3,6-TCN). For the PG7 pigment, five of these fragments were identified, with TCBN not being detected. Figure 3 illustrates the trends for the selected fragments, highlighting their behavior under the specified experimental conditions. To enhance clarity, the TeCBDN fragment, which showed higher intensity in all the graphs, is presented only in the inset. This approach allows the main plots to emphasize the remaining fragments without signal overlap, using a customized scale for each graph to achieve this separation.

Chromatographic peak area of selected fragments normalized to the internal standard, as a function of laser power at a fixed irradiation time (25 min), in A and C for the GC ink and PG7 pigment, respectively, and as a function of irradiation time at a fixed laser power (3 W), in B and D for the GC ink and PG7 pigment, respectively. The dots indicate fragments from the decomposition of chlorinated phthalocyanine, and the squares indicate those from the active chlorination of naphthalene impurities

The analysis of the GC samples exposed for 25 min to different power levels revealed a linear trend in the intensity of the fragmentation products as the applied power increased (Fig. 3A) for both classes of fragments. The observed linear behavior is supported by high values of the R2 coefficient obtained for each of the six fragments, indicating a strong correlation between the applied power and the production rate of the fragments. Specifically, the R2 values for the fragments are as follows: trichlorobenzonitrile (0.987), tetrachlorobenzonitrile (0.9814), tetrachlorobenzodinitrile (0.9902), pentachlorobenzonitrile (0.9824), 2,7-dichloronaphthalene (0.979), and 2,3,6-trichloronaphthalene (0.9588). Conversely, when the GC ink samples were treated at a constant power of 3 W for increasingly longer exposure times, no correlation was observed with the intensity trend of these fragments (Fig. 3B).

In contrast to the GC ink, which exhibited a strong linear correlation between all fragments and the applied laser power, the PG7 pigment showed no such dependence for most fragments (Fig. 3C). The lower R2 values for the PG7 pigment are likely attributable to its instability in dispersion. Notably, only the TeCBDN and 2,7-DCN fragments are correlated with the applied power, with R2 values of 0.8012 and 0.9114, respectively. Under constant power conditions, the time-dependent analysis of PG7 revealed that only the TeCBDN and PCBN fragments linearly increased over time, with corresponding R2 values of 0.7257 and 0.8769, respectively (Fig. 3D).

The analysis of the fragments indicates that TeCBDN is commonly identified in both pigment and ink samples following laser treatments, suggesting a direct association with the decomposition of the PG7 pigment. Furthermore, there is a noticeable dependence on the applied laser power for the production of this fragment in both samples.

4,5,6,7-Tetrachloro-2,3-dihydro-1H-isoindole is detected in all PG7 samples, whereas in the GC ink samples, it is only observed following treatment with high power or after prolonged exposure times, suggesting that its formation is occasionally below the limit of detection (LOD) of the GC‒MS method.

Given the observed linear increase in the production of all fragments with applied power in the GC ink analysis, the magnitude of this increase was evaluated for each of them. This was accomplished by calculating the production rate of all fragments via the normalized peak area at 3 W divided by that at 0.5 W. The results of this analysis are presented in Fig. 4.

Production rates of all the fragments, calculated for the GC ink samples treated for 25 min at 3 W and 0.5 W, along with their respective chemical structures. The 100% threshold indicates equal production rates at 3 W and 0.5 W

The production rate of all common fragments was calculated according to formula (1):

$$PR=\left(\frac_}_}\right)\%$$

(1)

where PR is the production rate and PA is the peak area for the 25-min treatments at 3 W and 0.5 W, respectively. These values are indicative of the dependency of fragment production on the pulse power in W.

The production rate quantifies the extent to which the intensity of a specific fragment’s production increases with applied power. A PR of 100% indicates that the fragment production rate at 3 W matches that at 0.5 W. Higher values indicate a percentage increase in the production rate under higher power conditions relative to 0.5 W treatment, which is taken as the reference.

Although all fragments exhibit some degree of increase, the TeCHI, TCBN, and TeCBDN fragments show the most pronounced increase in production as the applied power increases. These fragments maintain the integrity of the tetrachlorobenzene ring. Notably, tetrachloronaphthalene was consistently absent in our analysis and in previous studies of PG7 fragmentation, whereas naphthalene and methylnaphthalene have always been detected.

Principal component analysis (PCA) was conducted to provide further information on the relationship between fragmentation and laser power for the GC ink and PG7 pigment samples. The analysis was conducted on the normalized peak areas of the fragments from both the GC ink and the PG7 pigment, which were treated for 25 min with laser powers ranging from 0.5 to 3 W (Fig. 5). Two fragments were excluded from the analysis: TCBN, due to its absence in the PG7 samples, and TeCHI, because it was present only in the G25-1.25 and G25-3 samples.

PCA of the GC and PG7 samples treated for 25 min: score plot (left panel) and loading plot (right panel), including the percentage of explained variance on each axis. The blue dots correspond to the GC, the orange dots correspond to PG7, and the violet dots correspond to the PCA weights of the fragments

The analyses were conducted on normalized samples, which were centered and scaled to unit variance via the following equation:

$$Scaled X=\frac_ - mean\left(X\right)}$$

where Xi represents the peak areas of the fragments, mean(X) represents the mean value averaged over the peak areas of X fragments and sd(X) represents the standard deviation of the normalized peak areas. This ensures that features with large variances do not dominate the PCA representation.

The score plot of the PCA revealed that the first principal component (PC1) accounts for 70.4% of the total variance between the samples, whereas the second principal component (PC2) explains 21.9%.

PC1 predominantly differentiates the GC samples on the basis of the applied power, with the samples arranged in order of increasing laser power along the horizontal PC1 axis. In contrast, the PG7 samples exhibit a weaker separation along this axis. However, the distinction becomes more pronounced for the PG7 samples along the second component (PC2), although it is not linear, as previously observed, but rather, the difference between the two extremes of applied power is considered.

The position of the majority of the fragments in the loading plot shows that, compared with that of the PG7 pigment, the fragment intensity increases with increasing power applied to the GC ink.

The separation along PC2 for the PG7 pigment samples is also explained by the positions of fragments TeCF and PCBN in the loading plot. Compared with those of the GC samples, the intensities of these peaks indicate greater fragmentation of PG7 at 3 W of power. Furthermore, the TeCBDN fragment in the loading plot indicates that its intensity increases with increasing applied power for both sample types.

PCA confirmed a stronger dependence of fragment intensity on power in the GC samples than in the PG7 samples, as an increase in power led to a corresponding increase in fragment production. Previous semiquantitative analyses revealed that, for six selected fragments, this increase followed a linear trend. The positioning of TeCBDN in the loading plot suggests that its production increases in response to increasing power across both sample types.

SEM images were collected from samples deposited on silicon wafers (used as sample holders) immediately at the end of each treatment, and as such, snapshots of the particle sizes and morphologies related to the processes triggered by exposure to the femtosecond laser were obtained. Figure 6 shows images of the most common particles, aggregates, filaments and composite structures found in both the GC and PG7.

SEM images of representative samples treated with various powers and durations of irradiation with a Ti:sapphire femtosecond laser. GC ink was treated for A 5 min at 3 W and B, C, and D 30 min at 3 W (different areas). PG7 pigment treated for E and F 5 min at 3 W (different areas), G 20 min at 3 W, and H 30 min at 3 W

In general, femtosecond laser treatment causes processes of fragmentation and melting, similar, to some extent, to those of nano- and picosecond lasers. However, peculiarities can be observed, both in terms of the shape and size distribution of the treated samples although distinctions must be made between GC and PG7. The most striking feature of all the GC-treated samples was the dominant presence of particles or small aggregates in the range between 20 and 120 nm, an example of which is reported in Fig. 6A. Their regularity is quite distinctive, since it could not be observed, to this extent, on equivalent samples treated with nano- and picosecond lasers (Cecchetti et al. 2022; Bauer et al. 2020). Simultaneously, larger particles form as a result of a melting process, as shown in Fig. 6B, C, highlighted by orange arrows or a circle. The latter, in particular, indicates a melted area emerging from the initial sheath of particles embedded in the ink, which forms islands. Larger agglomerates, with a diameter on the order of tens of microns, also form, as shown in Fig. 6D, and are more abundant with increasing power.

Finally, piercing of the initial sheath by the laser beam is observed at several points, two of which are marked by yellow rings in Fig. 6C.

A quantitative assessment of the particle size distribution was performed through statistical analysis of areas of 1 mm2 from SEM images of each sample, and the results are reported in Fig. 7A–C. Figure 7A presents an example of the particle size distribution in sample G30-2.25, which is in the 20–300 nm range and indicates an increasing frequency with diameter, with a peak in the 96–110 nm range, followed by a frequency drop, for particles larger than 110 nm. Notably, nanoparticles embedded in the sheath were excluded from the statistics, and for asymmetric particles (i.e., those with an aspect ratio > 1), the largest dimension was selected for counting. In addition, nanoparticles in the 2–3 nm range are also observed, but in small amounts, they are on the order of 1 every 50 with sizes > 20 nm. They have also been excluded from general counting because their identification requires higher magnifications, which results in the loss of an overview and the absence of correct counting. To assess the overall size distribution as a function of power and time, 3D plots were generated, where the particles were grouped into three categories, i.e., with diameters < 200 nm, 201–500 nm and > 501 nm and labeled NPs, MPs and AGs, respectively. The plot of the size distributions at 25 min and increasing power is shown in Fig. 7B, whereas the histogram at 3 W and increasing time is shown in Fig. 7C. In these analyses, the points after 5 min of treatment were excluded because of the prevalence of untreated sheaths in the images. The subtotal is the sum of the NP and MP frequencies. The trend observed as a function of power indicated a frequency peak of the NPs at 2.25 W and, correspondingly, a minimum of MPs at the same power. On the other hand, the trend of the subtotal is a monotonic decrease to a plateau that parallels a monotonic increase in the AGs. This is in line with the decrease in the absorbance of the UV‒Vis spectra with increasing power to a plateau, suggesting that the generation of particles < 500 nm is responsible for the spectral variations. The trend of the frequencies at 3 W with time has up-and-down trends for all categories. However, in general, an increase in the frequency of the NPs corresponds to a decrease in the number of MPs, and in this case, the subtotal frequency parallels the intensity of the spectral variations in the corresponding UV‒Vis spectra.

Quantitative assessment of the particle size distribution via SEM images of the treated samples. A Particle size distributions of the GC samples treated for 30 min at 2.25 W. 3D plot of the particle size distributions of the GC samples: B treated for 25 min with increasing laser power and C treated at 3 W with increasing time. The particles are categorized on the basis of their diameter into NPs (< 200 nm), MPs (201–500 nm), AGs (> 501 nm), and subtotal particles (sum of NPs and MPs)

The SEM images obtained upon treatment of PG7 with the femtosecond laser reveal some peculiarities, largely related to the level of aggregation, which is much stronger than that of GC because of the inherent hydrophobicity of the pigment and its additives. This results in extended agglomerates, as shown in Fig. 6E, where beam piercing is also visible (marked by yellow rings). Islands also form in this case although they are larger than those in the GC. For comparison purposes, the average dimensions of the islands were estimated to be 2.5 ± 1.7 mm × 3.2 ± 2.1 mm for GC and 3.7 ± 2.0 mm × 8.3 ± 3.0 mm for PG7.

Agglomerates with a large aspect ratio (rods) form at the top power and intermediate exposure time (20 min, Fig. 6G, P20-3) and are similar to those observed by Kihara et al. (Kihara et al. 2019), who purposely grew long agglomerates upon nanosecond treatment followed by aging.

Most notably, thin fibers formed upon treatment of the PG7 at the top power (Fig. 6H, P30-3) and were present at all the different exposure times (whereas at the same level of treatment, the GCs displayed large agglomerates). Notably, a similar behavior, i.e., the appearance of fibers, was observed both upon Nd:YAG nanosecond laser treatment (3.5 J/cm2, 15 and 44 min) of the pigment PG36 (Bauer et al. 2021), which was dispersed either in water or in propan-2-ol, and from the extract of the GC, which was dispersed in water and subsequently treated with a Nd:YAG nanosecond laser (0.525 J/cm2, 120 min) (Bauer et al. 2020).

Nanoparticles ranging from 20 to 120 nm in size, as well as those on the order of 2–3 nm, are also observed, but with a far lower frequency than that of the GC. Owing to the large inhomogeneity of the formed particles as well as the large degree of agglomeration, a statistical analysis of the particles was not performed.