Introduction

The discovery of laser-induced surface relief grating (SRG) on azo polymer films nearly 30 years ago initiated enormous research into this intriguing effect1,2. The inscription of a variety of polymer topographic waviness ranging from one-dimensional sinusoidal or asymmetric to more complex two-dimensional circular or spiral ones has been demonstrated using light interference patterns or structured illumination1,2,3,4,5,6,7,8,9. The topographic modifications, being permanent below polymer glass transition temperature but erasable or reconfigurable when needed, showed remarkable application potential in optical data storage and nanolithography10,11.

Considering the microscopic processes involved, the effect of SRG formation arises from clean and reversible trans–cis photoisomerization reactions of the azobenzene derivatives10,12. Due to angularly dependent absorption of linearly polarized light by elongated trans azo molecules and numerous trans–cis-trans reactions, the molecules tend to align in directions perpendicular to the polarization direction of the excitation light. According to a modern orientation approach proposed by Saphiannikova et al., the chromophore reorientation leads to the rearrangement of the polymer backbones coupled to them and the appearance of mechanical stress13,14. A permanent macroscopic material deformation may occur when the stress is higher than the polymer yield stress.

In the class of glassy azo polymers, the formation of SRGs was studied in functionalized, supramolecular, and guest–host systems. Even though the distinct materials were prepared and exposed to different optical fields (i.e., different intensity and polarization patterns), some general rules regarding the efficiency or dynamics of the phenomenon have been quickly discovered. The rules referred, in particular, to the type of bonding between the azo moiety and the polymer chain, polymer molecular weight, azo chromophore content, or the polarization states of the interfering beams10,15,16,17,18.

Besides the details of an azo polymer structure and applied irradiation conditions, the polymer thickness has been demonstrated to affect the SRG inscription process. The polymer layers for the SRG formation studies have been prepared on glass substrates using spin-coating or casting techniques. Typically, the thickness of the spin-coated films is ca. 500 nm, as in the case when the SRG phenomenon was observed for the first time1,2. The height of the inscribed SRG was ca. 20% of the original film thickness as pointed out by Kim et al.2, but soon after, an over 1000 nm deep SRG in a 1200 nm-thick polymer layer was reported19. It was suggested by the authors that the AFM tip reached the glass substrates for some of the gratings as the flattened profile in troughs was noticed.

Investigating different azo polymers, researchers observed a general trend of faster growth of SRG amplitudes or larger diffraction efficiencies when using thicker samples, and some of the authors gave interesting hypotheses19,20,21,22,23. In one of the early studies, the grating formation effect was suggested to be a volume rather than a surface one as no differences in diffraction efficiencies of the SRGs produced in the samples with thickness above a certain value were found20. On the other hand, the hindering of the grating inscription in extremely thin films, regardless of light intensity, was attributed to the pinning effect of the substrate interface21.

Considering the above, the issue of layer thickness-SRG efficiency dependence becomes essential when comparing the behaviour of different materials. However, the approach to the problem appears not to be standardized. For example, Priimagi et al. prepared samples with varying thicknesses but similar optical density at the writing wavelength when studying the efficiency of SRGs in a series of hydrogen-bonded polymer-azobenzene complexes with different chromophore concentrations17. On the other hand, Yadavalli et al. used polymer films of the same thickness of 500 nm to perform comparative studies between a commercially available PAZO and two other azo polymers after finding the SRG amplitudes increasing from a few tens of nanometers in a 250 µm-thick PAZO sample to ca. 900 nm in a 1-µm-thick one22.

Recently, we have reported on an extraordinarily deep SRG of 1.75 µm generated in a custom-synthesized azo polymer (labeled as PI-Az) with an arbitrarily chosen thickness of 3.3 µm24. Having the material with a remarkable ability to form relief structures, we now use it to systematically study the layer thickness—SRG efficiency relationship, including the final achievable depths and details of SRG growth dynamics. We prepare the PI-Az layers with four different thicknesses, covering a wide range of optical density values at the recording wavelength, from almost entirely to merely transmissive, and apply prolonged exposure to the light interference pattern. The measured temporal evolutions of diffraction efficiencies during the grating recording were used to propose the mathematical model for the SRG formation and to compare the grating growth dynamics in the studied layers. The topography profiles along a few millimetre-long SRG diameters and the neighbouring scratches in the PI-Az layers were taken to show the mutual distance between the valleys of each SRG and the substrate surfaces. Finally, the Fourier analysis was applied to the whole profile of the sample with the largest modulation depth to calculate its diffraction pattern and compare it with the pattern observed experimentally.

Methods

The studied polymer material PI-Az was synthesized as reported recently25. The PI-Az layers were prepared by casting polymer solution (PI-Az in N-methyl-2-pyrrolidone solvent) onto clean glass substrates followed by drying at 120 °C overnight. The film thickness was measured with a Dektak stylus profiler while scanning the stylus across the scratched layer surfaces.

For the inscription of SRGs, a 442 nm light from a HeCd laser (Kimmon Koha) was used. The irradiation process was carried out in the experimental set-up presented in the previous work24. Briefly, two beams of right- and left-handed circular polarization crossed at 2.7°, forming the 9.3 µm-period interference pattern (according to the formula: Λ = λ/(2sin(θ/2)), where λ is the writing wavelength and θ is the intersection angle)18. The intensity of each Gaussian beam was ca. 100 mW/cm2 at the sample plane, where the circular beam’s diameter was 2.1 mm (1/e2). During irradiation, the process of a diffraction grating inscription was probed with the aid of an attenuated 690 nm beam focused to the size of ca. 100 µm in the centre of the interference region. The polarization of the probe beam was horizontal. Three silicon detectors placed behind the sample measured the optical power of the 690 nm beams diffracted into 0th, + 1st, and + 2 orders in 1-s intervals. Apart from the thinnest sample, the irradiation was carried out for ~ 10 h to verify whether the saturation of diffraction efficiency signals could be reached. Since the diffraction efficiency of the mth order of a thin sinusoidal phase grating is given by the square of the Bessel function of the first kind and mth order26:

$${J}_{m}^{2}\left(\frac{\Delta \varphi }{2}\right),$$
(1)

where: \(\Delta \varphi =\frac{2\pi }{\lambda }d\cdot (n-1)\) is the amplitude of the phase modulation of the passing light with wavelength λ, d is a peak-to-valley grating profile amplitude, n is the effective refractive index of the grating material (the air is the surrounding medium),

each set of experimental 0th, 1st and 2nd order diffraction efficiency signals was fitted using the \({J}_{m}^{2}(\frac{\Delta \varphi (t)}{2})\) functions with m = 0, 1 and 2, respectively.

After holographic recording, the sample’s topography profiles and diffraction properties along the illuminated spot diameter were examined. In the former examination, a Dektak XT stylus profiler was applied, while in the latter, the same focused probe beam as during the holographic recording process and a motorized translation stage with a sample holder were used. During the movement of the sample across the probe beam in a direction perpendicular to the grating grooves, the optical power of the produced diffracted beams was measured.

For the thickest PI-Az layer, the Fourier analysis was performed. The analysis was based on the entire topography profile (from a Dektak stylus profiler), which has been translated into the phase delay profile (assuming the refractive index of 1.75 and wavelength of 690 nm). The Fourier transform of such prepared input data was calculated and expressed as an angular spectrum. In this way, the number, positions, and efficiency of all occurring diffraction orders could be calculated. Moreover, for a better match with the experimental procedure, the illumination with a small laser dot was simulated by restricting the input data to a narrow range (corresponding to the diameter of the reconstructing laser beam). This active area can be chosen arbitrarily along the topography profile, which can be used to simulate the scanning along the sample. As the relief is the deepest in the center of the sample and utterly flat outside of the illuminated area, such an approach also gives an insight into the process of grating formation.

Results and discussion

The studied azo poly(ether imide) PI-Az contained one unsubstituted azobenzene chromophore per each repeating unit (Fig. 1)25. The azo moieties were rigidly attached to the main chain via one of the phenyl rings. As previously reported, the material was characterized by an amorphous structure, a high glass transition temperature of ca. 170 °C, and a relatively low molecular weight, indicating its oligomeric nature25. The UV–vis absorption of the PI-Az layer showed two characteristic absorption bands: a strong one at around 330 nm and a weak one at around 450 nm, attributed to the π-π* and n-π* electronic transitions of azo moieties, respectively25. The recording wavelength used (442 nm) was chosen because of its proximity to the maximum of the n-π* band, which allows, not only for an efficient transition from the trans to cis state but also from cis to trans.

Fig. 1
figure 1

Chemical structure of the studied azo poly(ether imide) – PI-Az.

Full size image

Table 1 presents the thickness of the prepared PI-Az layers, their initial transmittance, and optical density at the writing wavelength, together with the sample codes.

Table 1 Polymer thickness, initial transmittance, and optical density at the writing wavelength of the prepared samples, together with the sample codes.
Full size table

Since the initial transmittances at 442 nm of the PI-Az-1 and PI-Az-2 layers were 0.45 and 0.25, respectively, it could be assumed that the blue light intensity distribution throughout their entire thicknesses was almost uniform. The assumption was less valid for PI-Az-3, for which the initial transmittance was 0.11, and unreasonable in the case of PI-Az-4 where almost no blue laser light was observed behind the sample at the start of irradiation. Taking a criterion of the initial transmittance value and not the absolute polymer thickness, the PI-Az-1 and PI-Az-2 layers can be considered thin, while PI-Az-3 and PI-Az-4—thick.

The holographic recording was performed in experimental geometry characterized by a small crossing angle between the interfering beams to ensure the cross-section of the interference pattern lay in the plane of the polymer layer18,27. The resultant polarization state of two overlapping waves with counter-circular polarization is considered to be purely linear with the azimuth continuously rotating by 180 degrees over one modulation period. Using the model proposed by Xu et al.27 it was verified that in the case of illumination at an angle of 2.7° the actual polarization state deviated from the linear one by only 4.7% (the maximal eccentricity of the resultant elliptical polarization state was equal to 0.9989). The total intensity modulation did not exceed 0.5% and thus could be neglected.

It is worth noting that the writing configuration with such a small interference angle (offering a large grating period) has been rarely reported in the literature and was even shown disadvantageous for SRG production in a specific azo polymer18.

Solid lines in Fig. 2 present the measured temporal evolution of diffraction efficiencies of the 0th, + 1st, and + 2nd orders, (calculated as the ratio of the optical power of the 690 nm light diffracted in the particular order to the total optical power transmitted through the sample) for PI-Az-1, PI-Az-2, PI-Az-3, and PI-Az-4.

Fig. 2
figure 2

Diffraction efficiencies vs. exposure time for (a) PI-Az-1, (b) PI-Az-2, (c) PI-Az-3, and (d) PI-Az-4; solid lines-the measurement; dashed lines-the mathematical fit.

Full size image

The most striking feature arising from the comparison of the four data sets is the distinct shape of the curves, revealing a completely different grating inscription efficiency. In the case of PI-Az-1, holographic irradiation resulted only in a slight increase in the optical power of the + 1st diffraction order. Due to the poor efficiency of the grating formation process, the irradiation was stopped after ca. 1 h. Much more pronounced changes in the intensity of the 0th and 1st orders were observed for PI-Az-2. The + 1st diffraction order grew systematically up to 9%—efficiency value accompanied by an evident decrease of the 0th order signal. A few minutes after starting the irradiation, the continuously rising 2nd diffraction order was detected; however, its final value was only 0.4%. On the contrary, nonmonotonous characteristics with maxima and minima were seen for the two thickest samples (Fig. 2c-d). In the case of PI-Az-3, the intensity of the non-diffracted beam dropped to practically zero during a ca. 3-h recording and then began to increase with further irradiation. This was accompanied by the growth of the + 1st diffraction efficiency (within ca. 1.5 h) up to 31%, after which the signal started decreasing to a final 13%. The + 2nd-order diffracted beam was much more intense than in the case of PI-Az-2. For the thickest PI-Az-4 sample, all the curves were characterized by one fully—and another—partially resolved maxima.

The overall shape of the curves measured for PI-Az-3 and PI-Az-4 strongly indicated diffraction on a sinusoidal phase grating as it resembled the squares of the Bessel functions of the first kind. Nevertheless, the steep slopes of the 0th- and 1st-order curves at the initial irradiation times and the significant asymmetry in the peak curvatures suggested that the processes of the grating formation did not proceed linearly with time. To gain insight into the grating formation dynamics, curve fittings were performed using squares of the Bessel functions \({J}_{0}^{2}\left(\frac{\Delta \varphi \left(t\right)}{2}\right),{ J}_{1}^{2}\left(\frac{\Delta \varphi \left(t\right)}{2}\right),\) \({J}_{2}^{2}(\frac{\Delta \varphi (t)}{2})\) and a double-exponential growth for describing the temporal changes in the maximal light phase modulation:

$$\frac{\Delta \varphi }{2} = A\left( {1 - {\text{exp}}\left( { - \frac{t}{{\tau_{A} }}} \right)} \right) + B\left( {1 - {\text{exp}}\left( { - \frac{t}{{\tau_{B} }}} \right)} \right),$$
(2)

where A, B, and τA, τB represent the amplitudes and time constants related to the grating formation process. It should be noted that a model with a single exponential growth function did not reproduce the experimental curves correctly. The set of the fitted amplitudes and time constants are collected in Table 2 and the dashed lines in Fig. 2 represent the fitting curves.

Table 2 Fitting parameters for Eq. (1).
Full size table

Generally, an excellent agreement between the experimental and fitting curves proved the nonlinear grating inscription dynamics in each sample. We postulate that each exponential function is solely related to the relief formation process and represents its two stages. The hypothesis can be quickly confirmed by estimating the first-order diffraction efficiency due to birefringence grating. Taking the value of optical birefringence induced by the blue beam Δn \(\approx\) 0.016 from work28, the phase modulation amplitude of ca. 0.35 rad arises in the case of the thickest PI-Az-4 layer, which in turn leads to the first-order diffraction efficiency of only 1%. The bulk birefringence grating probably showed up as deviations between the experimental and fitting curves at initial irradiation time, nevertheless, including a short-time constant τC of the additional C-amplitude process in the proposed model led to meaningless fitting results.

The essential conclusions arising from the curve fitting are as follows:

  1. (i)

    The values of the τA and τB time constants do not show a strong dependence on the layer thickness.

  2. (ii)

    Both the A and B amplitudes increase evidently with polymer thickness, but a substantial rise is found for the latter parameter. As a consequence, while for the PI-Az-2 sample, both process stages have equal contribution to the total phase modulation (and the SRG depth), in the case of the two thickest samples the contribution of the B stage (i.e., B/(A + B)) almost reaches or exceeds 80%.

To investigate the possible origin of the two-stage SRG growth, prolonged holographic irradiation was repeated for another PI-Az layer with an initial transmittance at 442 nm equal to ca. 20%. This time, however, a single photodetector was placed close behind the sample to measure the optical power of the blue laser beams: the two transmitted 0th orders, and all the self-diffracted higher orders that were appearing with time. As shown in Fig. 3, the sample transmittance increased during irradiation with varying dynamics.

Fig. 3
figure 3

Temporal evolution of the PI-Az sample transmittance at 442 nm measured during holographic recording; the enlargement of the initial 200-s period in the inset.

Full size image

Two effects could contribute to the observed trend: a decrease in the absorption coefficient of a 442 nm light due to azo molecule reorientation and local decreases in sample thickness due to the SRG formation, which surpasses simultaneous local layer thickening. In the case of deep surface modulations, the local sample thickening could give rise to light-intensity modulation in the polymer plane at the SRG valleys. Thus, significant modifications of material internal structure and morphology might, in turn, affect the SRG inscription rate. The effect could be considered to correspond to a change in the initially applied irradiation conditions.

The expected light-induced modification of the azo polymer surfaces due to holographic recording was easily noticed during the inspection of the samples in daylight. While tilting the samples in different directions, the circular shining spot or a set of concentric dark or bright rings was visible to the naked eye in reflection (Fig. 4a). For the thinnest sample, a tiny shining spot was hardly seen at the centre of the illuminated region. The circular features in Fig. 4a arise from diffraction on the gratings with variable groove depths, generated by the non-homogeneous optical field having radial variation of the intensity profile (see text below). Each ring corresponds to the regions of the structures introducing the same phase retardation.

Fig. 4
figure 4

(a) Photographs of the sample surfaces after holographic recording observed under white light; PI-Az-2—on the left, PI-Az-3—in the middle, PI-Az-4—on the right. (b-d) The topography profiles taken along the intentionally made scratches (located at ca. x = 0) and along the whole holographically-exposed areas of the (b) PI-Az-2, (c) PI-Az-3, (d) PI-Az-4 surfaces; the arrow “SL” indicates the substrate level; the enlargement of the surface profiles in the central regions along a ca.100 µm distance in the insets.

Full size image

Using a stylus profiler, the surface profiles were measured across the whole illuminated regions, that is, along a few millimeter distances, in the direction perpendicularly to the grating grooves. Before taking the scans, scratches in the polymer layers down to the glass substrates were made in the vicinity of the illuminated regions to show the layer thicknesses. Figures 4b-d present the measured surface profiles for PI-Az-2, PI-Az-3, and PI-Az-4, respectively, with the enlargements of the central 100-µm fragments in the insets.

Three essential findings arise from the obtained scans. Firstly, all the SRGs exhibited a varying amplitude (along the x direction), being a consequence of the Gaussian profile of the interfering beams. The highest amplitudes of the surface deformations at the central locations were 0.16, 0.9, and 1.85 µm for PI-Az-2, PI-Az-3, and PI-Az-4, respectively. Secondly, the peaks and valleys of each SRG appeared quite symmetrically relative to the initial polymer surface level. Thirdly, in neither sample, notably, PI-Az-4, the deepest valleys of SRGs reached the glass substrate. On the contrary, for PI-Az-4 they were found 1.7 µm above the glass surface; simultaneously, the relief peaks were formed 0.85 µm above the initial sample surface.

Figure 5 presents the photographs of the diffraction patterns observed behind PI-Az-2, PI-AZ-3, and PI-Az-4 samples when a linearly polarized 690 nm beam was incident normally at the central locations of the inscribed SRGs.

Fig. 5
figure 5

Photographs of the diffraction patterns observed on the screen behind the (a) PI-Az-2, (b) PI-Az-3, (c) PI-Az-4 sample, when the 690 nm probe beam was incident at the central locations of the inscribed SRGs. The vertical dashed line indicates the 0th-order diffraction spots. Slight disproportions in the spot sizes and their mutual separation seen between the right and left part of the photos result from a side position of the camera.

Full size image

Each pattern was symmetrical with practically equal intensity of positive and negative diffraction orders. It was verified that the polarization azimuth of all the diffracted beams was identical to that of the incident beam. Moreover, no significant differences in the intensity between + 1st and -1st orders were noticed for a circularly polarized probe beam of either handedness incident on the edges of the relief structures. The result revealed that the unique birefringence grating possessing the optical axis that is a perfect replica of the rotating light polarization pattern was not induced in the polymer layers29. This could be due to a birefringence relaxation process evidenced in our recent study28, that destroyed the “cycloidal” azo chromophore alignment, and the sensitivity of the bulk birefringence grating to the polarization state of the probe light.

When the samples with the deepest SRG inscribed, i.e., PI-Az-3, and PI-Az-4, were translated against the 690 nm beam, the number and intensity of the diffracted beams were varied. Figure 6 presents the diffraction efficiencies of five positive diffraction modes for PI-Az-3 and PI-Az-4, measured using the same probe beam, as during the grating recording process. (The efficiencies presented in Fig. 6 are the relative ones, i.e., calculated as the ratio of the optical power of the light diffracted in the particular orders to the total transmitted optical power).

Fig. 6
figure 6

The diffraction efficiency plots versus the location x of the incident probe beam on the inscribed grating for (a) the PI-Az-3, (b) the PI-Az-4 sample; the polarization state of the probe beam was linear.

Full size image

Finally, the Fourier analysis was performed for the whole surface profile of the PI-Az-4 layer. Only the measured deformation profile was taken into account. Although the grating is formed within the volume of the polymer, it remains in the thin grating regime (Raman-Nath operation). This can be evaluated by calculating the Q parameter30,31:

$$\text{Q}=\frac{2\pi {\lambda }_{0}L}{{\Lambda }^{2}{n}_{0}},$$

where \({\lambda }_{0}\) is the vacuum wavelength of light, \(L\) is the thickness of the grating, \(\Lambda\) is the period of the grating, and \({n}_{0}\) is the mean refractive index. The highest amplitude of the surface deformations at the central location was \(L = 1.85 \mu {\text{m}}\). Assuming the wavelength \(\lambda =690 \text{nm}\), grating period \({\Lambda } = 9.3 \mu {\text{m}}\), and the refractive index \({n}_{0}=1.75\), we can calculate the Q parameter as Q = 0.053. It shows that the thin grating regime is fulfilled since Q < 1. Bragg regime operation is considered in the case of Q > 10. Therefore, even though the PI-Az-4 diffraction grating is almost three times thicker than the wavelength, it can be assumed to be thin grating in calculations and simulations performed further as Fourier analysis. Such a diffractive element can be understood as a higher-order kinoform32 that introduces a phase shift larger than 2π.

The constant refractive index equal to 1.75 was assumed, which means that the influence of the refractive index grating was neglected. Such an approach is reasonable as the influence of such a grating is roughly two orders of magnitude weaker than the physical deformation grating. The theoretical diffraction efficiencies of the first 6 diffraction orders produced by the obtained grating are shown in Fig. 7.

Fig. 7
figure 7

The theoretical diffraction efficiencies of the first 6 diffraction orders of the PI-Az-4 sample as a function of the spatial position of the reconstructing beam on the sample.

Full size image

Each data point in Fig. 7 was obtained by extracting the region having 100 µm width (which corresponds to the diameter of the reconstructing beam). This region of interest was moved across the sample in the steps of 20 µm. In every step, the Fourier analysis was performed to investigate the angular spectrum of the light behind the grating. The positions of the particular diffraction orders were determined (they coincided with the expected and measured ones), and the calculated intensity in these positions was given with respect to the whole transmitted power. In this way, the diffraction efficiencies from the 0th (central, non-bent beam) up to the 5th order of the diffraction were given. They are represented in Fig. 7 by subsequent lines labeled according to the legend.

As can be seen, the obtained characteristics are closely related to the measured ones (Fig. 6b). In both cases at the far edges of the sample, the whole radiation goes into the central 0th order of diffraction. In other words, there is no bending of the incoming beam, which is expected as no grating is imprinted into the sample at these positions. The sample acts as a simple plane-parallel plate. As the probing beam is moved towards the center of the sample (from both sides), one can notice the diminishing of the central beam and subsequent increase of the amount of the optical power transmitted into higher diffraction orders (firstly, the 1st one, secondly the 2nd one, etc.). It is connected with the increasing phase difference within the sample, as the investigated region lies more and more towards the center, where the strongest illumination occurred. From the diffraction point of view, the height of the grating is increasing, providing higher phase contrast between the valleys and the peaks of the inscribed grating. An interesting phenomenon occurs near the positions marked as 1 mm and 3.5 mm in Fig. 7, where the diffraction efficiency of the 0th order begins to rise again (at the cost of higher orders, for example, the 1st one). It shows, that the maximal difference in the phase delay (height) exceeded the 2π value. From these points onwards the center of the grating, the higher-order-kinoform regime of operation begins.

Overall, the high compliance of the registered intensity characteristics of all diffraction orders with the calculated one proves the credibility of the grating shape obtained from the profiler.

Conclusions

By performing prolonged holographic recordings in a series of azo polymer samples, we demonstrated the huge impact of polymer thickness on surface relief inscription efficiency. The inscription process in the azobenzene-containing amorphous material that exhibited oligomeric structure and efficient trans–cis cycling under the applied irradiation could be either carried out very efficiently or stopped entirely by varying the thickness of the polymer layer. The extraordinarily efficient inscription was found in the 2.7 µm-thick layer, strongly absorbing the incident light at the start of the irradiation process. However, achieving the final SRG amplitude as high as 1.85 µm required that the holographic exposure was extended above the time when the intensity of the beam diffracted from the grating into the first order approached its first theoretical maximum of 34%. Such an extension of the exposure time resulted in the detection of diffraction efficiency signal(s) oscillations.

On the other hand, the lack of relief formation was observed already in the layer, with 0.6 µm-thickness being highly transmissive to the incoming light. Such thickness definitely cannot be regarded as either ultrathin or thin. On the contrary, it is typically used in many studies on SRG fabrication.

The curve fitting performed to the data sets of diffraction efficiencies versus the exposure time showed that a biexponential growth function could model the SRG buildup. The parameters representing the amplitudes of both process stages increase significantly with the layer thickness, although the rise in the slower process amplitude is much more pronounced for thicker layers and contributes strongly to the SRG formation process.

The surface topography measurements (by stylus profiler) taken along the whole inscribed SRG areas and the nearby intentionally made scratches down to the substrates showed that in either of the azo polymer layers, the valleys of the SRG did not reach the glass plates. This finding and the saturating character of the relief formation dynamics indicate that this will never happen regardless of the irradiation period. Moreover, the above finding indicates that the glass substrate beneath the polymer layer does not block the relief growth.

Although the thickness of the polymer layer is essential for the relief inscription, considering its absolute value is insufficient for selecting the layer that offers an efficient SRG formation process. Our study indicates that instead, the layer optical transmittance at the excitation wavelength appears to be the critical parameter. Efficient relief generation was found in the samples with a low initial transmittance, i.e., a nonuniform initial light intensity distribution across their thickness. The observation indicates further that the original layer transmittance (or absorbance) should be considered when preparing the samples of different azo polymers for comparative studies of their capability of SRG formation.