1. Introduction

Lithium niobate (LiNbO₃, LN) crystals are extensively studied for their many important applications [1, 2], e.g., holographic volume storage, optical image and signal processing, coherent optical amplification, and phase conjugation. In particular, holographic volume storage has attracted increasing attention for the past three decades. One crucial problem with this application is the volatility of stored information, because during readout carriers are redistributed homogeneously, which leads to erasure of the recorded hologram. To solve this problem, two-color recording was proposed and has been the focus of research in recent years [3–10].

The two-color recording process in LN doubly doped with Fe and Mn was interpreted as a two-center picture (Fe and Mn centers) [3]. For nominally pure near-stoichiometric LN, this process was proved to be connected with the small polaron (the intermediate state) formed at the antisite defect Nb_Li [4–7]. The small polaron together with the bipolaron (electrons trapped at adjacent Nb_Li and Nb_Nb sites) plays a key role in the realization of two-color holography [5]. Recently, UV-light-induced absorption and two-color holography in LN highly doped with damage-resistant impurity have been demonstrated by many researchers [8–10]. Since Nb_Li has been eliminated completely in these samples, small polarons are excluded as the origin of these phenomena. The researchers attributed it to the creation of intermediate shallow centers O^- (i.e., bound small polarons), where UV-excited holes were trapped at O^2- sites near cation vacancies charged negatively with respect to the lattice [8–10]. In particular, Tomita et al. [10] investigated the UV-light-induced two-color photorefractivity in a near-stoichiometric LN doped with Mg and found that there are at least four types of centers participated in this two-color photorefractive effect, but the exact nature of all these centers was not given. For clarifying their essence, further investigations on LN samples with other damage-resistant impurity are needed. Recently, Kokanyan et al. found that doping of tetravalent elements Hf can substantively suppress the optical damage of LN in the visible spectral region [11]. Razzari et al. reported that the light-induced birefringence changes observed for LN doped with 4 mol% of Hf are comparable to those found for 6 mol% Mg-doped crystals and suggested that the so-called damage-resistant threshold was close to 4 mol% for Hf-doped LN [12]. Later, Li et al. gave the UV-visible and infrared absorption spectra of highly Hf-doped LN [13]. However, information is still lacking with regard to its optical properties in the UV spectral region.

In this paper we investigated the UV-light-induced absorption (ULIA) in LN highly doped with Mg and Hf, respectively and demonstrated their distinct decay behavior. The influence of the sample status (oxidized or reduced) on the ULIA was also studied. Based on these experimental results, the origin of the different centers participated in the ULIA process was discussed in detail.

2. Experimental procedure

Samples used in this study were LN doped with 6.5, 7.8 mol% MgO and 4.0, 6.0 mol% HfO₂, respectively. The highly doped LN single crystals were grown along the z axis from the congruent melt by using the Czochralski technique. The as-grown crystals were cut to rectangular-shaped Y-oriented plates, which were then polished to optical grade. To get the reduction state, the samples were treated in an argon atmosphere at 700°C for 6 h, and oxidation required treatment in air at 800°C for 10 h. The labels and material parameters of the samples are shown in Table 1.

Table 1. The labels and material parameters of the samples used in this study.

View Table | View all tables in this article

 

Fig. 1. Schematic of the experimental arrangement for the ULIA measurement. Details of uniform illumination of strong green light were omitted. UV light intensity was 500 mW/cm², intensity of strong green light was 200 mW/cm² , and that of the probe-light (denoted as “Pȁ) 1mW/cm². Another reference light (denoted as “R”) was used to reduce the drifts caused by power fluctuation of the semiconductor laser.

Download Full Size | PDF

Figure 1 shows a schematic experimental arrangement for the ULIA measurement. A 10 W mercury lamp was used as the incoherent UV light source. The spectrum of UV light was selected to peak at 365 nm by use of an appropriate optical filter. The UV light was loosely focused by a lens and irradiated the sample for pump. We used an e-polarized 532 nm light beam emitted from semiconductor laser as probe light, which impinged on the sample along the direction orthogonal to the Y plane of the sample. The transmitted light was detected by a photo-detector with a green filter placed in the front to block scattered 365 nm pump UV light. Additionally, another strong incoherent 532 nm light beam was expanded for uniformly illuminating the sample. Since the changes of the sample transmission during the ULIA process are very small (sometimes less than 1%), it is necessary to avoid external influences on the ULIA measurement. In our experiments the samples were kept at a steady temperature (295 K), and the absorption induced by the temperature change could be eliminated. In addition, another reference light was taken out from probe light before the sample. Also, the ULIA coefficient changes were obtained from ln[(I_R1/I_R0)/(I_P1/I_P0)]/d, where I_P1 (I_P0) and I_R1 (I_R0) are measured intensities of transmission and reference beams with (without) UV light illumination, respectively, and d is the sample’s thickness. This way we were able to reduce as much as possible the drifts caused by power fluctuation of the semiconductor laser. In order to test the reliability of our experiments, the absorption change of background was measured ahead for a long time. The absorption change was found to be nearly zero all the time, which indicates the external influences on the ULIA measurement had been effectively suppressed.

 

Fig. 2. Typical ULIA curve for highly Mg-doped LN crystal. ON- and OFF-states of UV light are denoted as arrows in the figure.

Download Full Size | PDF

3. Results and discussion

Figure 2 is the typical ULIA curve for a highly Mg-doped LN sample, which increases rapidly at the beginning of UV light irradiation, achieves a saturation value for several seconds, and decays partly to another stationary value in the dark after shutting down of UV light. It should be noted that “in the dark” mentioned here and later is only an approximation in the experimental condition. In other words, the samples are not situated in absolute darkness, even after the shutting down of UV light, because the green probe is still irradiating it in order to monitor the absorption. In our experiment, however, the probe light with the intensity of 1mW/cm² is very weak and can be neglected. This suggestion is supported by nearly the same results obtained when detecting the absorption from time to time instead of using a continuing probe. In Fig. 2, the most noticeable phenomenon is the nondecayed part of the ULIA in the dark, which can, however, be erased completely by uniformly strong illumination with 532 nm green light (as shown in Fig. 3). This result means that at least two type of centers participated in the ULIA process: one is unstable and decays in the dark, which corresponds to the shallow centers reported previously [8, 10]; the other can exist stably in the dark but is sensitive to the green light, which seems in agreement with the intermediate deep centers suggested by Tomita et al. [10]. The terms “shallow”, “intermediate deep”, and “deep” used here and later are concerning energy level for holes but not electrons. The complete decay of the ULIA for LN highly doped with Mg was reported by Zhang et al. [8], and the nondecayed part was not observed in their experiments. We think two factors should account for this discrepancy. First, higher intensity adopted for the green probe light may lead to the complete decay of the ULIA, because a relatively strong probe light can erase undesignedly the nondecayed part mentioned above. Another factor is the oxidation status of the sample used in experiments, which also influences the nondecayed part of the ULIA as discussed in the following paragraphs.

 

Fig. 3. Dark decay and green-light-induced decay curves for ULIA. ON-state of strong uniform illumination with 532-nm green light is denoted as arrows, black curves correspond to the experimental data, and red curves are the fitting results of the dark decay curves with bi-exponential form given by Eq. (2).

Download Full Size | PDF

 

Fig. 4. Experiment data for decayed and nondecayed parts of the ULIA in the dark.

Download Full Size | PDF

Figure 4 gives the detailed experiment data for decayed and nondecayed parts of the ULIA in the dark. First, we can see that the decayed part for Mg78a is obviously larger than Mg65a, which indicates Mg78a has more shallow centers than Mg65a. Assuming O^- near cation vacancy ($V_{\mathit{\text{Li}}}^{\mathit{-}}$ for LN) as the shallow center just like the previous suggestion [8–10], we can deduce more $V_{\mathit{\text{Li}}}^{-}$ exiting in Mg78a than Mg65a, which conflicts with the fact that the amount of $V_{\mathit{\text{Li}}}^{-}$ will decrease with Mg concentration above the so-called damage-resistant threshold [14]. Therefore, the origin of shallow centers should be considered again. We know, except $V_{\mathit{\text{Li}}}^{-}$ , highly Mg-doped LN has another type of defect: ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ . It is also charged negatively with respect to the lattice and may trap the holes at O^2- sites. Considering more and more ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ appears with the increase of Mg concentration, assuming O^- formed near ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ as the shallow center becomes reasonable. In the view of defect structure, O^- formed near $V_{\mathit{\text{Li}}}^{-}$ is more stable than ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ because of the loss of Li cation, and it may be corresponding to the intermediate deep center.

Another noticeable result comes from highly Hf-doped LN crystals. In contrast to highly Mg-doped LN, the dark decay of the ULIA was absent for both Hf40a and Hf60a, and only the similar green-light-induced decay curves can be observed. For simplicity, merely the experimental data of Hf60a are given in Figs. 2 and 3. This absence of the dark decay means the corresponding shallow centers have disappeared in these samples, which further weakens the possibility of O^- formed near $V_{\mathit{\text{Li}}}^{-}$ as the shallow center because of a large amount of $V_{\mathit{\text{Li}}}^{-}$ existing in LN highly doped with tetravalent hafnium ions [13]. Also, Xu et al. [15] investigated the ULIA for LN highly doped with trivalent In, and gave the much smaller dark decay as compared with LN highly doped with bivalent Mg. These results indicate that the shallow center is related to the valence of doped ions, which can be explained by our assumption about the shallow center in the previous paragraph. In LN highly doped with Mg, In, and Hf, the corresponding impurity defects formed at niobate sites are ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ , ${\mathit{\text{In}}}_{\mathit{\text{Nb}}}^{2-}$ ,and ${\mathit{\text{Hf}}}_{\mathit{\text{Nb}}}^{-}$ , respectively. For their decreasing electrical negativity, the ability of ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ , ${\mathit{\text{In}}}_{\mathit{\text{Nb}}}^{2-}$ , and ${\mathit{\text{Hf}}}_{\mathit{\text{Nb}}}^{-}$ for trapping holes goes down in turn. Thus, as the shallow centers formed in LN highly doped with Mg, In, and Hf reduce accordingly, so does the dark decay caused by them. For the extreme case of LN highly doped with Hf, the shallow centers responsible for the dark decay disappear completely.

In order to confirm our suggestion about the intermediate deep centers, we studied the influence of oxidation/reduction treatment on the ULIA of highly Mg-doped LN samples. Obviously, the treatment has no obvious effect on the dark decay of the ULIA, but it influences the nondecay part tremendously. From Fig. 3, we can see that the reduction process leads to the sharp decrease of the nondecay part. Generally, LN may loss oxygen near cation vacancies more easily than elsewhere during the reduction process. Thus, the amount of O^- formed near $V_{\mathit{\text{Li}}}^{-}$ will reduce after the reduction treatment, which causes the decrease of the intermediate deep centers.

Table 2. The fitting results of the dark decay curves in Fig. 3 using bi-exponential form given by Eq. (2).

View Table | View all tables in this article

For the further analysis of shallow centers, we fitted the dark decay curves for all samples. These curves cannot be well described by the functional form used by Zhang et al. [8]:

but can be perfectly fitted with bi-exponential form given as following:

where A₀ denotes the nondecay part of ULIA.

 

Fig. 5. Structural schematics of lattice for highly Mg-doped LN. ${\mathit{\text{Mg}}}_{\mathit{\text{Li}}}^{+}$ is located near ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ for the balance of electric charge, but it doesn’t participate in the charge transfer process of ULIA.

Download Full Size | PDF

Tomita et al. [10] also approximated the dark-decay trend of the shallow-center grating to an exponential form rather than from Eq. (1) and gave a good fit to their data. Table 2 lists our fitting results. We can see the values of A₀ in this table have the similar trend as given in Fig. 4. In addition, all samples have nearly identical τ₁ (about 9 s) and τ₂ (about 100 s), which indicates the existence of two different dark decay processes, A ₁exp(-t/τ ₁) and A ₂ exp(-t/τ ₂) . In nature, the dark decay can be interpreted as the relaxation process of holes from shallow traps to deep ones in the dark. However, the origin of deep traps in highly Mg-doped LN remains disputable. These traps should be quite close to the conduction band but have an ability to supply holes (be able to trap electrons). Some researchers [10, 16] considered ${\mathit{\text{Fe}}}_{\mathit{\text{Nb}}}^{2-}$ as the hole-supplying deep center, but the direct evidence of ${\mathit{\text{Fe}}}_{\mathit{\text{Nb}}}^{2-}$ close to the conduction band is absent until now. The broad band centered near 0.9–1 eV (1.1–1.3 μm) was always observed in reduced highly Mg-doped LN and attributed to small polaron absorption via polaron hopping at Nb_Nb sites [17–19]. It indicates that ${\mathit{\text{Nb}}}_{\mathit{\text{Nb}}}^{5+}$ has an ability to trap electrons (supply holes), can form a level relatively close to the conduction band, and perhaps play the role of the hole-supplying deep centers in highly Mg-doped LN. This suggestion is also supported by the fact that holes have been found by ESR (Electron Spin Resonance) to be created together with electrons trapped at ${\mathit{\text{Nb}}}_{\mathit{\text{Nb}}}^{5+}$ (forming Nb⁴⁺) [20]. Now, we tentatively take ${\mathit{\text{Nb}}}_{\mathit{\text{Nb}}}^{5+}$ as the hole-supplying deep centers and give a description of one possible mechanism for ULIA decay in highly Mg-doped LN. UV light excites holes from ${\mathit{\text{Nb}}}_{\mathit{\text{Nb}}}^{5+}$ (the deep centers) to the valence band. After their migration in the valence band, part of them are trapped by O^2- near $V_{\mathit{\text{Li}}}^{-}$ and ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ , which leads to the creation of O^- and the corresponding ULIA. After the UV light is shut down, the ULIA caused by stable O^- near $V_{\mathit{\text{Li}}}^{-}$ (the intermediate deep centers) remain nondecayed, but the holes at the relatively unstable O^-near ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ (the shallow centers) relax rapidly to neighborhood ${\mathit{\text{Nb}}}_{\mathit{\text{Nb}}}^{4+}$ (the deep centers), which corresponds to the dark decay process of ULIA. Figure 5 shows a hypothetical model for ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ surrounded by two nearest Nb_Nb (six nearest Nb_Nb in all because of the threefold symmetry of LN). Due to the presence of a spontaneous polarization P_s in LN, Nb_Nb A and B are not equivalent in the view of energy. In comparison with A, holes trapped near ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ relax to Nb_Nb B more easily, which leads to two dark decay processes with different time constants (τ₁ and τ₂). Here, traps (for holes) at Nb_Nb A, B and O^2- near ${\mathit{\text{Mg}}}_{\mathit{\text{Nb}}}^{3-}$ constitute a three-level band scheme. Recently, Qiao et al. [21] suggested another three-level model to interpret their ULIA results for highly Zn-doped LN. However, using such a model to explain our experimental results is difficult. Just as they emphasized, more experimental support and further investigation are of great necessity to clarify the mechanism of the ULIA for highly doped LN.

4. Conclusion

The ULIA in LN highly doped with Mg and Hf was investigated, respectively. Both dark decay and nondecay parts of ULIA were observed in highly Mg-doped LN, but only a nondecay part was observed in Hf-doped LN. This distinct behavior was attributed to different centers (the shallow centers and the intermediate deep centers for trapping holes) formed under UV illumination. O^- near doped cation at the niobium site was suggested to be the origin of the shallow center, which is responsible for the dark decay and disappears in Hf-doped LN due to the very weak electrical negativity of ${\mathit{\text{Hf}}}_{\mathit{\text{Nb}}}^{-}$ . Meanwhile, O^- near cation vacancy corresponds to the intermediate deep center for the nondecay part. The influence of the sample status (oxidized or reduced) on the ULIA was demonstrated to support our augment. In addition, two different dark decay processes were associated with relaxations of holes from the shallow centers to two unequivalent Nb_Nb adjacent to the doped cations at niobium sites.

As a tetravalent ion, Hf has different photorefractive properties from Mg²⁺, Zn²⁺ , or In³⁺. At the time this paper was being written, Li et al. [22] found that Fe ions remain at Li sites in Hf- and Fe-codoped LN crystals when the HfO₂ doping concentration goes above its threshold value; as a result the photorefractive response rate and sensitivity are greatly improved, and the saturation diffraction efficiency remains at a high value. The different ULIA behavior of highly Mg- and Hf-doped LN observed in this work helped us to clarify the nature of different centers formed under UV illumination and reveal the detailed kinetics of the ULIA process involved in UV light-gating nondestructive two-color holography. For highly Hf-doped LN, the absence of the dark decay process means that no shallow centers form under UV illumination. It implies that the migration speed of holes in highly Hf-doped LN should be faster than that in highly Mg-doped crystal, which is important for improving the response speed of UV photorefractive holographic storage in LN.