1. Introduction

Lithium niobate (LiNbO₃, LN) is one of the most important synthetic crystals because of its excellent electro-optic, acousto-optic, elasto-optic, piezoelectric, pyroelectric, and nonlinear optical properties. like a decathlon winner, LN performs well in nearly all aspects [1], particularly in photonic applications, such as waveguide modulators, lasers, frequency conversion and photorefractive devices. Laser-induced optical damage (also named photorefraction) was discovered in LN and LiTaO₃ crystals [2]. and can be utilized for holographic storage, one of the next generation storage technologies. Photorefraction can be enhanced by doping with transition-metal ions, such as Fe, Mn, Cu, Ce [3,4]. Up to now, iron doped LN, (LN:Fe) has become the mainstream research material for volume holographic data storage for its high diffraction efficiency, high data-storage density, and long dark storage time. However, there are still two main deficiencies, such as the low response speed and volatility. Buse et al. achieved nonvolatile two-color recording based on LN doubly doped with iron and manganese (LN:Fe,Mn) [5]. However, the response time of LN:Fe,Mn, being of the order of minutes, remains too slow for practical use. At present, low response speed has become the main obstacle for the application of LN to realize a real-time read-write memory.

Laser-induced optical damage of congruent LN can be strongly suppressed by optical damage resistant additives, such as Mg, Zn, In, Sc, Hf, Zr [6–10], as soon as the doping concentration exceeds a certain threshold. Doping above this threshold also considerably improves the response speed [11]. E.g., the responsetime was shortened from minutes to seconds for iron doped LN by codoping with Mg, Hf or Zr [12]. UV holographic storage is nearly impossible in LN:Fe codoped crystals because of strong absorption. Recently, we have found hexavalent molybdenum doped lithium niobate crystals (LN:Mo) to permit fast response and multi-wavelength holographic storage [13]. These excellent characteristics were attributed to Mo⁶⁺ ions occupying regular niobium sites because of their high valency state. They produce the new defect type ${\text{Mo}}_{Nb}^{+}$. At a doping level of 0.5 mol%, the response time becomes as small as 0.35 s while still a high saturation diffraction efficiency of about 60% can be achieved at 351 nm. However, the response time in the visible range is still several seconds. In this paper, we investigated the holographic properties of LN:Mo codoped with Zr (LN:Mo,Zr) or Mg (LN:Mo,Mg).

2. Experiments

According to our former work [13], optimum results were achieved at a doping concentration of 0.5 mol% Mo and by a polarization current of 145 mA. For the present investigation, a series of congruent LN crystals doped with 0.5mol% Mo and codoped with different concentrations of Zr or Mg were grown along the c axis with the conventional Czochralski method. The [Li]/[Nb] composition was selected as 48.38/51.62. The doping concentration of ZrO₂ is 1.0 and 2.5 mol% and of MgO is 3.0 and 6.5 mol%, labeled as LN:Mo,Zr_1.0, LN:Mo,Zr_2.5, LN:Mo,Mg_3.0 and LN:Mo,Mg_6.5, respectively. The as-grown crystals were annealed at 1150°C for 10 h in ambient atmosphere and artificially poled under an electric current of 145 mA for 15 min at 1170°C. At last, 3.0 mm and 1.0 mm thick y-oriented plates were cut and polished to optical grade.

Holographic characteristics of these samples in the UV-visible range were investigated by two-wave mixing in transmission geometry at the wavelenghts 351, 488, 532, and 671 nm provided by an Ar⁺ laser and a cw frequency-doubled solid-state lasers, respectively. An extraordinarily polarized laser beam was split into two beams of equal intensity (intensity per beam being 238, 400, 400, and 1500 mW/cm², respectively). Two mutually coherent beams irradiated these 3.0 mm thick plates with a crossing angle of 30°. The grating vector was aligned along the c-axis to exploit the largest electro-optic coefficient r₃₃. The diffraction efficiency is defined here as $\eta =I_d/(I_d+I_t)$, where I_d and I_t are the diffracted and transmitted intensity of the readout beam, respectively. The photorefractive response time constant τ_r and the saturation diffraction efficiency η_s are deduced by fitting the function $\eta (t)={\eta }_s{[1-\exp (-t/{\tau }_r)]}^2$ to the data. According to Kogelnik’s coupled wave theory [14], we could calculate the amplitude of the refractive index change Δn through the formula $\Delta \text{n}=[\lambda \cos \theta /\left(\pi \text{d}\right)]\text{arc}\sin \sqrt{{\eta }_{\text{s}}}$, where λ is the wavelength in vacuum, θ is the half crossing angle of the two writing beams in crystal, d is the grating thickness (i.e., the thickness of the crystal in our case). The photorefractive sensitivity S was defined as $S={(d\sqrt{\eta }/dt)}_{t=0}/(IL)$, where I is the total recording light intensity and L is the crystal thickness.

OH⁻ spectra and the optical absorption spectra were measured on 1.0-mm-thick y plates using a Magna-560 Fourier transform IR spectrophotometer and a Beckman DU-8B spectrophotometer, respectively, with light transmitting along the y axis at room temperature. The change of absorption within the wavelength range 300–800 nm was obtained by subtracting the absorption of a CLN sample from that of the investigated samples.

3. Results and discussion

Typical curves of the diffraction efficiency as a function of time are shown in Fig. 1 for LN:Mo,Mg_6.5 at the wavelengths 351 nm (a), 488 nm (b), 532 nm (c), and 671 nm (d), respectively. The photorefractive properties of LN:Mo,Zr and LN:Mo,Mg samples are shown in Fig. 2. For comparison, the data of 0.5 mol% MoO₃ doped LN (LN:Mo_0.5) polarized with the same current of 145 mA, is also presented. Holographic storage is achieved from UV to the visible for all LN:Mo samples codoped with Zr or Mg. For LN:Mo,Zr, the diffraction efficiency increases when the light wavelength varies from 671 to 488 nm, but decreases at 351 nm, especially for LN:Mo,Zr_2.5. The response time is in the order of tens of seconds or even minutes in the visible range but shorter at 351 nm. Compared to LN:Mo,Zr_1.0 and

 

Fig. 1 The measured diffraction efficiency as a function of time for LN:Mo,Mg_6.5 at various wavelength. (a), (b), (c) and (d) are for 351, 488, 532, and 671 nm, respectively.

Download Full Size | PDF

 

Fig. 2 The photorefractive (a) diffraction efficiency, (b) response time, (c) refractive index change and (d) sensitivity of LN:Mo crystals codoped with different concentration of Zr (solid symbols) and Mg (open symbols) from UV to the visible. The light intensity per beam is 238, 400, 400, and 1500 mW/cm² for 351, 488, 532, and 671 nm laser, respectively.

Download Full Size | PDF

LN:Mo,Zr_2.5, even the response speed increases as the concentration of ZrO₂ increases, but is still slower than for LN:Mo_0.5. The refractive index change and sensitivity decrease when the doping level of ZrO₂ is improved. However, the situation is different for LN:Mo,Mg: the diffraction efficiency increases while the response time shortens when the wavelength varies from 671 to 351 nm. The diffraction efficiency of LN:Mo,Mg_3.0 is about 15 times higher than the one of LN:Mo,Zr_2.5 while the response time is close to one of LN:Mo_0.5 except at 671 nm. If the doping level of MgO is enhanced to 6.5 mol% and thus exceeds the threshold, LN:Mo,Mg_6.5 maintains the short response time of 0.22 s but has a higher diffraction efficiency of 67.3% compared LN:Mo_0.5 at 351 nm. It was reported that the UV response time of LN can be shortened to less than 0.1s by doping 9.0 mol% MgO [15]. But the data was measured at 325 nm, which is shorter than 351 nm and most importantly very close to the absorption edge of LN. Besides, the growth of LN:Mg crystal is very difficult when the concentration is as high as 9.0 mol%. It should be pointed out that the space-charge field and the response time are strongly dependent on the grating spacing, it is not suitable to simply compare the diffraction efficiencies and the response times at different wavelengths, for example, the grating spacing at 671 nm is approximately twice as compared with that at 351 nm. More surprising is that the response time of LN:Mo,Mg_6.5 is shortened to only 0.33 s and 0.37 s at 488 nm and 532 nm, respectively. This is about one order of magnitude shorter than for as-grown LN:Fe,Zr crystals, which were reported to be the fastest codoped LN:Fe crystals in visible range [12]. The refractive index change and sensitivity decrease when 3.0 mol% MgO is doped in LN:Mo_0.5, and then increase as the doping level of MgO is improved to 6.5 mol%. Especially LN:Mo,Mg_6.5 obtains a large refractive index change of 3.46 × 10⁻⁵ and a high sensitivity of 12.06 cm/J in the UV region.

Because Mo⁶⁺ ions push regular Nb⁵⁺ to Li sites, forming new defects of ${\text{Mo}}_{Nb}^{+}$ and a larger amount of anti-site Nb⁵⁺ (${\text{Nb}}_{Li}^{5+}$), so if all of the ${\text{Nb}}_{Li}^{5+}$ wanted to be cleared up, the real threshold of ZrO₂ and MgO in LN:Mo should be higher than their mono-doping values of 2.0 mol% and 4.6 mol%. It is well known that when the doping concentration of optical damage resistant dopants exceed their threshold, the OH⁻ absorption band shifts from the position at 3484 cm⁻¹ of pure LN to higher wavenumbers [9,10]. The OH⁻ spectra of our crystals are shown in Fig. 3. Obviously, the peak of LN:Mo,Mg_6.5 shifts to 3532 cm⁻¹ which is related with the formation of a $M{g}_{Nb}^{2+}-O{H}^{-}$ complex [17]. However, the peak shifts of LN:Mo,Zr crystals are not obvious. So a threepeaks-fitting model [18,19] was employed. The fitting results show that when the concentration of ZrO₂ reaches 2.5 mol%, the 3467 cm⁻¹ absorption peak vanishes and a new one appears at 3498 cm⁻¹. Therefore, these results indicate 2.5 mol% ZrO₂ and 6.5 mol% MgO are both exceeds their real threshold in co-doped LN:Mo.

 

Fig. 3 The OH⁻ Spectroscopy of LN:Mo,Zr and LN:Mo,Mg crystals with different doping level.

Download Full Size | PDF

The fact that LN:Mo,Mg_6.5 responds faster and that the diffraction efficiency is lower than for LN:Mo_0.5 in the visible range can be explained as follows: A large amount of ${\text{Nb}}_{Li}^{5+}$ caused by doping Mo into LN, induces more intrinsic photorefractive centers, such as small polarons (${\text{Nb}}_{Li}^{4+}$) and bipolarons (${\text{Nb}}_{Li}^{4+}:{\text{Nb}}_{Nb}^{4+}$) [20]. While moving in the conduction band of congruent LN:Mo crystals, electrons may be trapped by these intrinsic photorefractive centers. Therefore the average time of each excitation, movement, and trapping cycle is strongly increased, resulting in a low response in holographic recording. If Mg ions are doped into LN:Mo with concentrations above the threshold, these Mg ions will substitute ${\text{Nb}}_{Li}^{5+}$ ions completely. Therefore small polarons and bipolarons mostly disappear. Consequently, the covered distances as well as the lifetimes of excited electrons become longerand the concentration of donors is reduced. Therefore the response speed of LN:Mo,Mg_6.5 for visible holographic storage is significantly improved while the diffraction efficiency is decreased compared to LN:Mo_0.5.

However, it is strange that Zr⁴⁺ ions do not improve the response of LN:Mo such as Mg²⁺ ions above their threshold concentration. A possible reason may be the different valency of Zr and Mg ions. As Mo⁴⁺, Mo⁵⁺ and Mo⁶⁺ ions exist in our crystals [13], the lower valence of Mg²⁺ ions make them more likely occupy ${\text{Nb}}_{Li}^{4+}$ than is the case for Zr⁴⁺ ions. Similar to what happens in LN:Fe,Mg [11], if the concentration of ZrO₂ exceeds its doping threshold, the site occupancy of partial Mo⁴⁺ ions does not change. On the other part, just as for LN:Fe,Zr [12], codoping with ZrO₂ may change the ratio of Mo⁴⁺, Mo⁵⁺ and Mo⁶⁺ ions, which may reduce the amount of novel ${\text{Mo}}_{Nb}^{+}$ defects and thus weaken photorefraction. Nevertheless, these assumptions need further investigations.

As shown in Fig. 3, even though the doping concentrations of ZrO₂ and MgO both exceed their threshold, there is still big difference between the response time of them. Besides, the diffraction efficiency of LN:Mo,Zr_2.5 and LN:Mo,Mg_6.5 is almost the same in the visible but very different in the UV region. It is well known that many kinds of defects may serve as photorefractive centers and determine the photorefractive properties of LN. Therefore the nature of the photorefractive centers has to be clarified for our crystals. To that purpose, we first investigated the light erasing behavior of photorefractive gratings in the LN:Mo,Zr and LN:Mo,Mg. Here we just give the normalized typical erasing curves of LN:Mo,Mg_6.5 at 351nm for example, as shown in Fig. 4. The erasing curves exhibit large deviations when fitted by a mono-exponential function, but can be modeled quite well by a sum of two exponentials, i.e., by the fitting model $\eta ={\eta }_1\exp (-t/{\tau }_1)+{\eta }_2\exp (-t/{\tau }_2)$, where η_1,2 is the initial diffraction efficiency and τ_1,2 is the time constant, respectively. This implies that at least two kinds of photorefractive centers are involved.

 

Fig. 4 The normalized erasing curves of LN:Mo,Mg_6.5 at 351nm. (a) and (b) are curves fitted by mono-exponential function and double-exponential functions, respectively.

Download Full Size | PDF

Figure 5(a) shows the UV-visible absorption difference between our codoped samples and CLN. A pronounced absorption peak and a wide absorption region from 400 to 800 nm are observed for all crystals. The wide range of the continuous absorption spectrum means that charge carriers can be excited from different energy levels at various wavelengths. This explains why holographic storage can be realized for our samples from UV to the visible range.

 

Fig. 5 The absorption difference of LN:Mo,Zr and LN:Mo,Mg relative to CLN: (a) LN:Mo,Zr and LN:Mo,Mg with various doping concentration of Zr and Mg, (b) The fitting curve of LN:Mo,Mg_6.5, where the fitting trough and three peaks centered at about 316, 326, 337, and 480 nm, respectively.

Download Full Size | PDF

However, a UV trough appears for LN:Mo,Mg_6.5 that is not observed for other crystals. Fitting the absorption spectrum of LN:Mo,Mg_6.5 to a sum of Lorentz functions revealed a UV trough and three peaks centered at about 316, 326, 337, and 480 nm, respectively, as shown in Fig. 5(b). The UV photorefractive centers of doping LN is considered to be related to some defects, such as ${\text{O}}^{\text{2}-/-}$ near $V_{Li}^{-}$, $M{g}_{Nb}^{3-}$ and ${\text{Zr}}_{Nb}^{-}$ [21,22]. ${\text{Mo}}_{Nb}^{+}$ may also serve as UV photorefractive center [13]. Because the ${\text{Mo}}_{Nb}^{+}$ defect center has a positive valence of + 1, it may attract an electron from O²⁻. Thus, O⁻ polaron could be formed and the strong bond between molybdenum and oxygen ions will stabilize the O⁻ polaron. The OH⁻ spectra have shown that the concentration of 6.5 mol% MgO exceeds the threshold, which causes nearly all of the ${\text{Nb}}_{Li}^{5+}$ dismisses, thereby the concentration of $V_{Li}^{-}$ for keeping the electric charge equilibrium decrease, so the UV trough at 316 nm should be correlated with the $O^{2-/-}-V_{Li}^{-}$ defect. The UV peak at 337 nm corresponds to ${\text{O}}^{2-/-}-M{o}_{Nb}^{+}$ [13], and the peak at 326 nm may correspond to ${\text{O}}^{2-/-}-M{g}_{Nb}^{3-}$ defect cluster.

Though the OH⁻ spectra also show that the concentration of ZrO₂ exceeds the threshold, there is a distinctly difference, namely that LN:Mo,Zr_2.5 has much weaker UV photorefraction compared to LN:Mo,Mg_6.5. For the photorefractive center of ${\text{O}}^{2-/-}-M{o}_{Nb}^{+}$ exist in all of the codoped crystals, the difference is that $M{g}_{Nb}^{3-}$ has a higher negative valence than ${\text{Zr}}_{Nb}^{-}$, O²⁻ ions near $M{g}_{Nb}^{3-}$ are less stable and lose electrons more easily under the illumination of UV light than near ${\text{Zr}}_{Nb}^{-}$ [22]. Thereby, the difference in the capacity of ${\text{O}}^{2-}-M{g}_{Nb}^{3-}$ and ${\text{O}}^{2-}-Z{r}_{Nb}^{-}$ as donors may lead to the discrepancy in UV photorefraction between LN:Mo,Zr_2.5 and LN:Mo,Mg_6.5.

It is impossible to simply attribute the wide absorption band centered at 480 nm to bipolarons, for ${\text{Nb}}_{Li}^{5+}$ has been cleared up in LN:Mo,Mg_6.5. As the results of [13], the band was great influenced by the polarization process. Thus, the combination of bipolarons and other defects caused by the process of polarization, such as the defects related to oxygen, may lead to the band. The broad band is so complex that further detailed research should be done to clarify it.

4. Conclusions

In summary, a series of LN:Mo,Zr and LN:Mo,Mg crystals with different Zr and Mg doping were grown and characterized. Holographic storage from UV to the visible is realized in all of the doubly doped crystals. It is interesting that ZrO₂ cannot improve the response speed of LN:Mo even when its concentration is above the threshold. However, when the concentration of MgO exceeds the threshold, a very short photorefractive response time of 0.22 s, 0.33 s, 0.37 s and 1.2 s for 351, 488, 532 and 671 nm was obtained, respectively. Our experimental results indicate that Mg²⁺ ions are a preferable choice to improve the photorefractive response of LN:Mo crystal. LN:Mo,Mg can be an excellent candidate for all-color holographic data storage with fast response.