Enhanced Power Conversion Efficiency of Perovskite Solar Cells with an Up-Conversion Material of Er3+-Yb3+-Li+ Tri-doped TiO2

In this paper, Er3+-Yb3+-Li+ tri-doped TiO2 (UC-TiO2) was prepared by an addition of Li+ to Er3+-Yb3+ co-doped TiO2. The UC-TiO2 presented an enhanced up-conversion emission compared with Er3+-Yb3+ co-doped TiO2. The UC-TiO2 was applied to the perovskite solar cells. The power conversion efficiency (PCE) of the solar cells without UC-TiO2 was 14.0%, while the PCE of the solar cells with UC-TiO2 was increased to 16.5%, which presented an increase of 19%. The results suggested that UC-TiO2 is an effective up-conversion material. And this study provided a route to expand the spectral absorption of perovskite solar cells from visible light to near-infrared using up-conversion materials. Electronic supplementary material The online version of this article (10.1186/s11671-018-2545-y) contains supplementary material, which is available to authorized users.


Background
Organolead halide perovskite solar cells (PSCs) have become attractive in the solar cell field, which is due to their advantages, such as high efficiency, lost cost, and simple fabrication [1][2][3][4]. In a few years, the power conversion efficiency (PCE) of PSCs has been improved to 22.1% [5]. However, perovskite solar cells only absorb a small fraction of incident light in UV and visible ranges due to the narrow energy band gap of perovskite sensitizer; thus, a large portion of incident light is lost due to its non-absorption of near-infrared (NIR) [6,7].
One promising method to solve the NIR energy loss issue is to apply up-conversion materials to PSCs, which can convert NIR to visible light. Some authors have reported the applications of up-conversion materials to perovskite solar cells [8][9][10], in which the up-conversion materials adopted were mainly based on beta-phase sodium yttrium fluoride (β-NaYF 4 ). While the β-NaYF 4 up-conversion materials can reduce charge transport ability of electron transfer layer [11]. It has been reported that Er 3+ -Yb 3+ -F − tri-doped TiO 2 can improve the PCE of dye-sensitized solar cells (DSSCs) due to its enhanced up-conversion emission compared with Er 3+ -Yb 3+ codoped TiO 2 [12]. In our previous publication [13], we reported the application of Er 3+ -Yb 3+ co-doped TiO 2 nanorods to PSCs. Some researchers have proved that the addition of Li + into Er 3+ -Yb 3+ co-doped TiO 2 could increase the up-conversion emission [14,15]. And it has been reported that the perovskite solar cells based on Lidoped TiO 2 produce higher performances compared to the device based on un-doped TiO 2 [16]. Thus, we wonder if the up-conversion materials of Er 3+ -Yb 3+ -Li + tridoped TiO 2 can be applied to PSCs to further improve the performance.
Therefore, in the present study, we prepared Er 3+ -Yb 3 + -Li + tri-doped TiO 2 (UC-TiO 2 ) by addition of Li + into Er 3+ -Yb 3+ co-doped TiO 2 , which presented an enhanced up-conversion emission compared with Er 3+ -Yb 3+ codoped TiO 2 . The UC-TiO 2 was applied to perovskite solar cells. The PCE of the solar cells with UC-TiO 2 is increased to 16.5 from 14.0% for the solar cells without UC-TiO 2 , which presents an increase of 19%.

Fabrication of Perovskite Solar Cells
Patterned FTO glass substrate was cleaned in acetone, 2-propanol, and ethanol by sonication for 20 min, respectively. Then, UV-O 3 was used to treat the FTO for 15 min. A compact layer was formed by spincoating a precursor solution on FTO and annealed at 500°C for 30 min. The precursor solution is 0.1 M titanium diisopropoxide bis (acetylacetonate) (75 wt% in isopropanol, Aldrich) solution in 1-butanol. A mesoporous TiO 2 film was obtained by spin-coating TiO 2 solution on the compact layer at 4000 rpm for 30 s, followed by annealing at 100°C for 10 min and

Results and Discussion
Up-conversion emissions were measured with an excitation of a 980-nm laser. Figure 1a shows the up-conversion emissions spectra of Er 3+ -Yb 3+ -Li + tridoped TiO 2 (Er:Yb:Li:Ti = 0.5:10:x:100, x = 0, 10, 15, 20, and 25, molar ratio). Figure 1b shows [15,16]. With the increase of Li + doping content, the intensity of the spectrum increases firstly, and then decreases, which is the maximum when the doping content of Li + is x = 20.
Hereinafter, the up-conversion material of Er 3+ -Yb 3 + -Li + tri-doped TiO 2 (Er:Yb:Li:Ti = 0.5:10:20:100, molar ratio) was applied. To confirm the doping of Er, Yb, and Li into TiO 2 , XPS spectra of UC-TiO 2 were recorded and shown in Fig. 3. The peaks at 458.1 and 463.9 eV in Fig. 3a could belong to Ti 2p 3/2 and Ti 2p 1/2 , respectively, and the peaks at 168.8 eV in Fig. 3b and at 192.7 eV in Fig. 3c could be attributed to Er 4d and Yb 4d, respectively [18]. The peak at 55.5 eV in Fig. 3d can correspond to Li 1s [19]. The survey XPS spectrum of UC-TiO 2 and O1s peak were also presented in Additional file 1: Figure S1. The results demonstrated that Er, Yb, and Li atoms were doped into TiO 2 . Figure 4a displays the UV-vis-NIR absorption spectra of TiO 2 (30NR-D) and UC-TiO 2 . Compared with 30NR-D, UC-TiO 2 presents a stronger absorption, especially at the range from 900 to 1000 nm. The energy band gap a b (E g ) could be estimated with the Tauc plot [20]. The Tauc plots are shown in Fig. 4b, from which the values of E g could be obtained to be 3.20 and 3.10 eV for 30NR-D and UC-TiO 2 , respectively. The E g of UC-TiO 2 is smaller than that of un-doped TiO 2 . Figure 5a shows the SEM image of 30NR-D film formed on the compact layer. The nanoparticle size is about 30 nm, and the size distribution is uniform. Figure 5b shows the SEM image of the film containing 30NR-D and UC-TiO 2 deposited on the compact layer by spin-coating method. There is no obvious difference between the two films, which displays that the particle size and morphology of UC-TiO 2 are similar to those of 30NR-D.
In the present work, the perovskite film was formed with the method previously reported [17]. According to report, the composition of the perovskite film is Cs 5 (MA 0.17 FA 0.83 ) 95 Pb(I 0.83 Br 0.17 ) 3 , and the role of the CsI is to make the perovskite solar cells thermally more stable, with less phase impurities, and less sensitive to processing conditions [17]. The scheme of the device is presented in Additional file 1: Figure S2.
The perovskite solar cells based on mesoporous layer formed with the mixture of UC-TiO 2 sol and diluted TiO 2 solution (UC-TiO 2 :TiO 2 = x:100, v/v, x = 0, 10, 20, 30, and 40) were fabricated and their I-V curves were measured. The photovoltaic parameters were obtained from the I-V measurements. Figure 6a shows the PCE dependence of solar cells on the content of UC-TiO 2 (UC-TiO 2 :TiO 2 = x:100, v/v) in the mixture. With the increase of UC-TiO 2 content, the power conversion efficiency (PCE) of solar cells increases firstly, and then decreases, which is the maximum at the content of x = 20 for UC-TiO 2 . The detailed photovoltaic parameters of solar cells with 20% UC-TiO 2 and without UC-TiO 2 were listed in Table 1. Compared with those of the devices without UC-TiO 2 , the photovoltaic parameters of the solar cells with UC-TiO 2 present an improvement. The PCE of the solar cells with 20% UC-TiO 2 is increased to 16.5 from 14.0% for the solar cells without UC-TiO 2 , which presents an increase of 19%. Figure 6b displays the I-V curves of the typical solar cells with UC-TiO 2 and without UC-TiO 2 .
To understand the enhancement, some investigations were carried out. Steady-state photoluminescence (PL) and time-resolved photoluminescence (TRPL) can be applied to investigate the electron extration and transport process. The PL of perovskite layer on the mesoporous layers formed by 30NR-D and 30NR-D with UC-TiO 2 were measured and shown in Fig. 7a. Compared with that of 30NR-D/ perovskite, the PL intensity of 30NR-D with UC-TiO 2 /perovskite becomes reduced, which indicates that the electron extration and transport efficiency across the interface between 30NR-D with UC-TiO 2 and perovskite is better than that between 30NR-D and perovskite [21]. Figure 7b shows the TRPL spectra of perovskite layer on the mesoporous layers formed by 30NR-D and 30NR-D with UC-TiO 2 . The TRPL spectrum was fitted to a biexponential function, in which the fast decay (τ 1 ) may be resulted from transportation of free carriers, and the slow decay (τ 2 ) can be originated from radiative recombination of free carriers [22][23][24]. The obtained parameters are listed in Table 2. Compared with that of 30NR-D/perovskite, the fast decay time (2.8 ns) of 30NR-D with UC-TiO 2 /perovskite becomes smaller, while the fraction of fast decay process (98. 2%) becomes larger. This implies that the charge transfer between perovskite and 30NR-D with UC-TiO 2 is faster than that between perovskite and 30NR-D. Eletrochemical impedance spectroscopy (EIS) is an effective method to get some information on carrier transfer behavior. Figure 8a displays the Nyquist plots of the devices based on mesoporous layers formed by 30NR-D and 30NR-D with UC-TiO 2 , in which two arcs were observed. The arc at high-frequency could be resulted from the contact resistance between the interfaces, and the arc at low-frequency could come from the recombination resistance (R rec ) and chemical capacitance (C μ ) of the device [25,26]. The EIS was fitted with an equivalent circuit shown in Fig. 8b, and the obtained parameters are listed in Table 3. The series resistance of the devices based on 30NR-D with UC-TiO 2 becomes smaller than that of the devices on based on 30NR-D, while the recombination resistance of the former becomes larger than that of the later. This indicates the charge recombination was decreased and the charge transport was improved for the device based on 30NR-D with UC-TiO 2 .
To further prove the effect of UC-TiO 2 on photocurrents of the devices, the I-V curves of the devices based on the mesoporous layers without UC-TiO 2 and with UC-TiO 2 were measured under the simulated solar radiation in the wavelength range of λ ≥ 980 nm with a NIR filter, which are shown in Additional file 1: Figure S3. Compared with that of the device without UC-TiO 2 , the photocurrent of the device with UC-TiO 2 was obviously enhanced, which demonstrates that the incorporation of UC-TiO 2 in the device can transform the NIR light into visible light, which can be absorbed by the devices to generate photocurrent.
To explain the increased open circuit voltage (V oc ) of the solar cells, the energy band arrangements of UC-TiO 2 , TiO 2 , perovskite, and Spiro-OMeTAD were shown in Fig. 9 based on the absorption spectra (Fig.  4) and the literatures [27,28]. The conduction band edge of the UC-TiO 2 is lower than that of TiO 2 (30NR-D) due to its smaller energy band gap; thus, the conduction band offset between UC-TiO 2 and perovskite is larger than that between TiO 2 and perovskite. This could be one of the reasons to have a higher open circuit voltage for UC-TiO 2 based solar cells [29,30].
In summary, the PCE increase of the solar cells based on the mesoporous layer with UC-TiO 2 is due to the enlarged I sc and increased V oc . The enlarged I sc could be due to expansion of spectral absorption to near-infrared (NIR) range with up-conversion material, reduced recombination, and fast charge transfer of photo-generated carriers. The increased V oc may be attributed to the enlarged conduction band offset between UC-TiO 2 and perovskite.

Conclusions
Er 3+ -Yb 3+ -Li + tri-doped TiO 2 (UC-TiO 2 ) was prepared by addition of Li + into Er 3+ -Yb 3+ co-doped TiO 2 , which presented an enhanced up-conversion emission. The UC-TiO 2 was applied to the perovskite solar cells. The performance of solar cells with UC-TiO 2 was improved compared with that of control device. The I sc , V oc , and FF of solar cells with UC-TiO 2 were increased to 22. 2 mA/cm 2 , 1.05 V, and 70.8% from 21.0 mA/cm 2 , 1.01 V,

Additional file
Additional file 1: Figure S1. a Survey XPS spectrum of UC-TiO2. b O 1s peak. Figure

Availability of Data and Materials
All data are fully available without restriction.