Double-N doping: a new discovery about N-doped TiO₂ applied in dye-sensitized solar cells

Abstract

In this paper, we first investigated the optimal amount of ammonia to add, using ammonia as the single dopant for use in dye-sensitized solar cells (DSSCs). Using this optimal amount of ammonia, urea was introduced as the second N dopant. The DSSCs produced by double-N doped samples combine the advantages of increased visible light absorption with ammonia as the nitrogen dopant and enlarged interface area with urea as the second nitrogen dopant. Not only was V_oc increased, but J_sc was also enhanced. We observed that the double-N doped sample forms a new microstructure with more mesopores, which enhance the transfer of electrolyte. Because these mesopores are the combination core of generated electrons and holes, they need to be kept in delicate balance. When urea is brought in as the double-N dopant, the doped amount of N atoms was improved. As a result, η is increased to 7.58%, a 14% improvement compared with single-N doped TiO₂.

---

1. Introduction

Dye-sensitized solar cells (DSSCs)¹ have attracted increasing interest since they were reported by Grätzel and co-workers because of their excellent photo-electrical conversion efficiency and the potential to replace commercial solar cells based on silicon.^2,3 Many groups worldwide are researching ways to improve DSSC efficiency. In DSSCs, the photoanode serves as the core part: it determines photo-generated-electron transport and dye adsorption; thus, it influences photo-electrical conversion efficiency of DSSC to a great extent.⁴ TiO₂ is the most commonly used photoanode material because of its low cost, non-toxicity, superior photoelectricity, dielectric effect and photo-electrochemical stability. However, as a photoanode, TiO₂ suffers from some disadvantages: (1) it has a large band gap; and (2) it has oxygen vacancies in its crystal lattice. In the electronic transmission process, these oxygen vacancies will become the recombination sites of electron and hole pairs that can reduce oxidation state of dye molecules and electrolyte. As a consequence, photo-electrical conversion efficiency falls. To improve the situation, many methods have been introduced, including doping with some metal and non-metal elements such as Zn²⁺, Fe³⁺, and N⁻.^5–8 Among those, N-doped TiO₂ can produce the most significant effect.

Since studied by Asahi et al. in 2001,⁹ various methods have been tried for the N-doping of TiO₂ and remarkable effects have been achieved. Meanwhile, various nitrogen dopants such as aqueous ammonia, ammonium chloride, urea, and tri-ethylamine have been studied. But until now, research on N doping has focused on single nitrogen dopants. Different nitrogen dopants produced distinguishable effects because of their disparate structures. Because different nitrogen sources might also have some complementary action, we researched two nitrogen dopants.

In this paper, to investigate the effect of double dopants, different kinds of nano-crystalline N-doped TiO₂ electrodes were synthesized using ammonia as a single-N dopant and combinations of ammonia and urea as double-N dopants, using wet methods. The result shows that N-doped TiO₂ powders could be produced with different crystallite sizes, microstructure, surface areas, band gap and N-doping amounts. In addition, the photo-electrochemical properties of DSSCs assembled by TiO₂ films also differ. Double nitrogen dopants could improve the efficiency of DSSCs.

2. Experimental

2.1. Synthesis of different TiO₂ nano-crystalline particles

Different TiO₂ particles were synthesized by a wet method as follows:¹⁰ butyl titanate was added dropwise to a predetermined amount of ammonia water under vigorous stirring in an ice bath. One and a half hours later, a white precipitate was obtained. The precipitate was centrifugally washed with ethanol, distilled water, and ethanol successively. A white powder was recovered by drying the precipitate at 80 °C. After finishing the drying process, the powder was calcined in air at 500 °C for 4 h (5 °C min⁻¹); a yellow powder resulted. Through varying the amounts of ammonia water, the optimal concentration was determined. We also used distilled water instead of ammonia water, under the conditions and other processes described above. The particle obtained was designated as TiO₂. Furthermore, if urea was first added to the ammonia water, pale yellow particles resulted that were designated double-N/TiO₂

2.2. Fabrication of TiO₂ photoanode and DSSCs assembly

TiO₂ pastes were prepared as followed: 1 g powder and 0.5 g ethyl cellulose (EC) were added to the mixture of 30 mL ethanol and 4.06 g terpineol. Then we added 60 g zirconia beads to the mixture and ground it for 3 hours in a ball mill to get a suspension slurry. We obtained the TiO₂ pastes after evaporating the solvent.

Next, the paste was screen-printed in 6 layers on fluorine-doped tin oxide (FTO) conductive glass to form TiO₂ films. The N-doped or pure TiO₂ films were sintered at 500 °C (5 °C min⁻¹ warming; 325 °C, 375 °C, 450 °C each maintained for 5 min) for 15 min in air. After cooling, the films were immersed in 40 mM TiCl₄ at 70 °C for 30 min and sintered at 500 °C for 30 min (5 °C min⁻¹). After being cooled to 110 °C, the film was put into a dye solution of N719 for 24 h. The DSSCs were assembled with N719/TiO₂ film, a Pt counter-electrode and an electrolyte to form a sandwich-type cell. The injected electrolyte consisted of 0.6 M DMPII, 0.03 M I₂, 0.5 M 4-TBP and 0.1 M GuSCN in acetonitrile and valeronitrile (at a volume ratio 85/15).

2.3. Characterization of the TiO₂ nanocrystals

The phase structure of samples was identified by X-ray diffraction (XRD, D/MAX-2500, Japan, Rigaku) using Cu Kα radiation, a 40 mA tube current, 40 kV voltage and a 2θ range from 10 to 90. The microstructures of powders and the size of particles were observed by SEM (S-4800, Japan Hitachi Ltd). The surface area of samples was measured by the Brunauer–Emmett–Teller (BET) method (Tristar 3000, Micromeritics, USA). The substitution of the oxygen sites with nitrogen atoms in TiO₂ and the N doping amount were confirmed by X-ray photoemission spectroscopy (XPS, PHI-1600, PE, USA) equipped with Mg Kα (1253.6 eV) as the excitation light source and power at 300.0 W. The spot size was 0.8 mm². Afterwards, the band gap of powders, calculated by UV-vis spectra, were obtained by a UV-vis spectrophotometer (Shimadzu UV-180 0 in 10 mm quartz cell spectrometer).

2.4. Photovoltaic measurements of DSSCs

The amount of dye absorbed by DSSCs was measured by UV-vis. The process was as follows:¹¹ the dye was desorbed from TiO₂ films in a mixed solution of ethanol and water (volume ratio = 1 : 1), then the concentration set at 0.1 mol L⁻¹ by the addition of NaOH. We calculated the adsorption amount of dye by the absorption spectrum of desorption solution.

Photocurrent–photovoltage (I–V) curves were measured under simulated AM 1.5 irradiation (100 mW cm⁻²) and the photocurrent–voltage (J–V) characteristics were recorded on the Keithley 2400 Source meter (solar AAA simulator, oriel China, calibrated using a standard crystalline silicon solar cell). The electrochemical impedance measurements (EIS) of the photoanode were obtained using an electrochemical workstation (CHI-660D, Shanghai Brilliance Co., Ltd).

3. Results and discussion

Fig. 1 and Table 1 show that 3-N/TiO₂ can be regarded as optimal on the basis of a series of properties. It enhances not only the short-circuit current (J_sc) but also the open circuit voltage (V_oc). The V_oc values increased with the increased amount of improved N-doping because the V_fb of N/TiO₂ was shifted to the negative by the formation of an O–Ti–N band. The J_sc values first increased as the energy level of TiO₂ changed by the N doping, resulting in the improvement of the charge transport at the interface of TiO₂/dye and then decreased as the phase and domain interfaces produced in the doped layer hindered the charge transport.¹² Ma et al.¹³ concluded that photocurrent significantly rose but that electron lifetime decreased as the amount of N-dopants increased. Hence, the amount of dopants has an optimum value. In our research, this optimum was denoted as N/TiO₂.

Fig. 1 (a) I–V curves of N-doped photoelectrodes with different ammonia amounts; (b) the curves of efficiency changing with the amount of ammonia.

Table 1 Electro-chemical characteristic of DSSCs

Electrodes	V oc/mV	J sc/mA cm^−2	FF	η/%
1-N/TiO2	0.78	9.90	69.88	5.39
2-N/TiO2	0.78	11.00	66.31	5.69
3-N/TiO2	0.82	11.10	72.89	6.64
4-N/TiO2	0.80	10.22	69.90	5.72

---

Fig. 2 shows the XRD patterns of samples. XRD is the one of most important research techniques used to identify the phase of a material and internal structure information such as atomic or molecular structure. The anatase-characteristic diffraction 2θ peaks are at 52.325, 37.841 and 48.074. The rutile characteristic diffraction 2θ peaks are at 27.459, 36.104 and 54.364. The sample spectra show that the all four samples had anatase phase, while P25 showed diffraction peaks characteristic of the rutile phase. The figure also shows that the intensities of the diffraction peaks of double-N/TiO₂ and N/TiO₂ were lower than that of pure TiO₂, and at the same time, the widths of the diffraction peaks were different for the different samples. The difference could result from crystal fragmentation, a low degree of crystallinity and misalignment in the crystal. However, the above phenomenon proves the formation of new O–Ti–N bonds. We calculated the crystal size of the samples by Scherrer's equation¹⁴D = kλx/βcos θ and the XRD data: D represents the crystal size, k represents the dimensionless shape factor (0.89), λ denotes the wavelength of the X-ray irradiation with 0.1541 nm for Cu Kα radiation, and β expresses the half peak width; finally, θ is Bragg's diffraction angle. The calculated particle sizes, shown in Table 2 (19.85 nm, 16.53 nm, 12.8 nm, 12.5 nm), are consistent with the SEM observations. The SEM images in Fig. 3 illustrate that there are differences in microstructure between single-N doping and double-N doping. Double-N doped TiO₂ shows more irregular mesopores, which is of benefit for the transfer of electrolyte. These mesopores contribute to more surface area, leading to more dye adsorption; this is also one reason for the improvement in J_sc values. On the other hand, because electrons transfer to the bottom electrode through the mesoporous semiconductor network, these mesopores may become the site for the combination of electrons and holes.

Fig. 2 X-ray diffractogram of samples.

Table 2 Characteristics of powders

Samples	Crystallite size/nm	Surface area/m^2 g^−1	N/%	Band gap/eV
P25	19.85	53.65	0	3.26
TiO2	16.53	40.40	0	3.18
N/TiO2	11.5	79.69	2.6	3.23
Double-N/TiO2	12.8	94.59	3.3	3.15

---

Fig. 3 SEM micrographs of the powders: (a) N/TiO₂ (b) double-N/TiO₂.

As seen in Fig. 4(a), the UV-Vis spectra of P25, and single and double-N/TiO₂ powders are mostly similar. Nevertheless, there is obvious absorption at 400–500 nm for N-doped TiO₂ powder, which is a characteristic absorption of nitrogen. The cause of the visible light absorption has been debated. Giamello et al.¹⁵ suggested that the absorption in the visible light region of N-doped samples resulted from the N_b centers that contained either diamagnetic (N_b⁻) or paramagnetic (N^*_b) bulk centers. Serpone et al.¹⁶ reported that the reason for the absorption in visible light was the formation of color centers. Ma et al.¹⁷ concluded that the activity in the visible range arose because of the new state induced by the nitrogen doping that lay close to the valence band edge. Fig. 4(b) is the plot of the (vale²versus the hν).¹⁸ The E_g of TiO₂ powder can be calculated by the extrapolation method (α = 0). Table 2 lists the E_g of the samples. That the N element is doped successfully into TiO₂ was proven using XRD and UV-vis. To further reveal the difference between single-N and double-N, we investigated the N-dopant amount by XPS, which can not only measure the doped amount but also can measure the combined energy of the bond.^18–23Fig. 5 shows the combined energy peak of N 1s for two samples of (a) single-N and (b) double-N doped titania powder. Di Valentin et al. proposed that N-doped TiO₂ has two structures:^21,22 in one, N was substituted for O at a regular lattice site; and the other had N as the host in an interstitial position directly linked to lattice O. The binding energy of the two kinds of N atom was different: two N 1s peaks of single-N doped arise at about 399.4 and 401.9, whereas two N 1s peaks of double-N doped lie at about 399.4 and 402.1. According to the binding energy table, N 1s binding energy in a N–Ti–N bond is 397.3; N 1s binding energy at about 402 is γ-N₂.²⁰ N 1s peaks in published studies lie mostly in the range of 396 to 404 eV.^18–22 In some cases the peak at 397.3 shifts to a higher binding energy, which is at about 399 eV. Burda et al. explained this phenomenon by the binding energy of the N 1s peak in O–Ti–N being higher than that in the N–Ti–N bond. Hence the peak at 399 may be the N in the O–Ti–N bond.²⁴ Compared with that of single-N doped samples, the intensity of the N 1s peak of double-N doped sample is stronger, meaning that more O–Ti–N was generated. The N-doped amounts of single-N and double-N doped titania powder are 2.6% and 3.3%, respectively; this indicates that the double-N doping can improve η by the two nitrogen dopants influencing each other. The two peaks of the Ti 2p are separated by about 5.7 eV from the Ti⁴⁺ peak. The peaks of Ti 2p_1/2 of the two samples are 530.15 and 529.75, corresponding to that of the Ti⁴⁺ oxidation state. Compared with single-N doped samples, the peak of the Ti 2p of double-N doped samples shifts 0.4 eV to a lower BE because more Ti³⁺ is produced. Oxygen vacancy increases with rising Ti³⁺; thus, in the bottom of the conduction band, it will introduce a shallower level as an electron trap to separate electrons and holes. The peak of O 1s is 529.9 eV, corresponding to the Ti–O bond.²⁵

Fig. 4 (a) UV-vis spectra of samples (b) the plot of (αhν)²versus photon energy (hν)).

Fig. 5 N 1s (a, single) (b, double) Ti 2p (c) and O 1s (d) XPS spectrum of single-N and double-N doped titania powder.

Sato et al.²⁶ suspected that titanium hydroxide generated by butyl titanate hydrolysis exhibited acidity, which was designated as titanic acid. The titanic acid reacted to form NH₄⁺. This doping progress can be described as follows:

As the reaction proceeds, NH₄⁺ in the system is constantly consumed. As a result, the concentration of NH₄⁺ decreases and the driving force of reaction is insufficient. But due to the introduction of urea, whose hydrolysis process can continue to produce NH₃, the driving force with which nitrogen was doped into the lattice enhances.

Fig. 6 shows that the current–voltage curves of DSSC equipped with N-doped and undoped TiO₂ photoanodes. The values of V_oc, J_sc, fill factor (FF), photoelectric conversion efficiency (η) and amount of dye adsorbance are listed in Table 2. Compared with pure TiO₂, V_oc is obviously improved, increasing by nearly 5%. It can be concluded from eqn (1) that the Fermi level of TiO₂ changed after N-doping:

V_oc = |V_fb − V_red| (1)

Fig. 6 I–V curves of N-doped and undoped photoelectrodes.

In eqn (1), V_red represents the reduction potential of a redox couple. Therefore, V_red should be constant in a set of parallel experiments. The improvement of the value of V_oc indicates that the Fermi level of TiO₂ has changed. This conclusion is consistent with that reported by Haujun Tian.²⁷ Compared with pure TiO₂, the value of J_sc had increased significantly. The amount of dye adsorbance is an important factor influencing J_sc. This may result from the N-doping changing the energy level of TiO₂, which reduced the value between the intrinsic Fermi level of material and lowest unoccupied molecular orbital (LOMO) of N719. Therefore, the loss of the photoelectrons from the dye-excited state into the conduction band of TiO₂ decreased. Lastly, the increase of photoelectric conversion efficiency also is attributed to the replacement of part of the oxygen atoms by nitrogen atoms to form the O–Ti–N structure. This effectively prevented the charge from recombining on the TiO₂–dye–electrolyte interface.

EIS was originally a method for measuring network frequency response in a linear circuit. Through its development, it became a powerful method to characterize the internal resistances performance of DSSCs. Its Nyquist diagram consists of three semicircles that represent three values of impedance.²⁸ The first semicircle shows the resistance of the interface between the electrolyte and Pt electrode, and the resistance of TiO₂/FTO; the second shows the resistance on contact surface between the TiO₂ dye and electrolyte; the third shows the resistance of electrolyte transportation. Fig. 7 shows the Nyquist, Bode plot diagrams and the electron lifetime of three samples. The fitting legend is indicated by colors as shown at the top right corner of picture (a). It can be seen from Fig. 7(a), the second semicircle, that the decreased resistance of TiO₂/dye/electrolyte was due to the electron transfer from I₃⁻ to the HOMO of dye and from the LUMO of the dye to the conductive band of TiO₂ becoming faster.²⁹ There is no obvious difference between single-doped and double-doped on the second semicircle. This may mean that there is no obviously increase on transmission speed after double-N doping. To some extent, it can be assumed that the double-N doping changed the conductive band a bit, leading to a driving force of electron injection that cannot increase electron transport speed more noticeably. Combined with the BET, the change of the third semicircle can be explained by TiO₂ not only having a larger surface area to uptake more dye molecules and more mesopores to transfer electrolyte after N-doping.

Fig. 7 (a) EIS spectra of N-doped DSSC photoelectrodes. (b) Bode plots.

Fig. 7(b) shows the corresponding EIS Bode plots. The peak frequency of the first arc represents the lifetime of electrons at the TiO₂–dye–electrolyte interface. It is clear from this set of plots that the lifetime of electrons is N/TiO₂ < double-N/TiO₂ ≈ TiO₂ because the existence of the mesopores restricts the electron transmission path to a greater extent.³⁰ Nevertheless, these mesopores of the double-N/TiO₂ sample are shown to be smaller, which reduces their adverse influence. In addition, the mesopores become the center of electronic capture, which is the reason that the lifetime of electrons decreases.³¹

The Mott–Schottky plot was obtained by a DSSC filled with an electrolyte solution contained 0.6 M DMPII, 0.03 M I₂, 0.5 M 4-TBP and 0.1 M GuSCN in acetonitrile and valeronitrile (at a volume ratio of 85/15), where a Pt counter electrode, a saturated calomel reference electrode (SCE) and a working electrode were immersed. The Mott–Schottky results were calculated as the inverse square of an apparent interfacial capacitance C as function of potential E, with: C = −1/2πfZ′′ (f is the test frequency, Z′′ represents the imaginary component of the interfacial impedance³²). The intersection point of the tangent and horizontal axis is the flat-band potential (V_fb). From Fig. 8, it can be concluded that all samples are n-type semiconductors. In addition, we can observe a negative shift in the flat-band potential with the doped samples versus pure TiO₂. This value for Double-N doped TiO₂ is −0.79. It is noteworthy that the doping changes the Fermi level and the flat-band potential of the TiO₂ (Fig. 9).

Fig. 8 Mott–Schottky plot of DSSC equipped with different samples.

Fig. 9 The UV-vis of a DSCC photoanode equipped with different samples.

According to Beer's law:

A = KCL

In this equation, A is the absorbance at a particular wavelength (for N719, 515 nm); K is a constant at 515 nm (1.41 × 10⁻⁴ dm³ mol⁻¹ cm⁻¹); L is the width of cuvette. Importantly, we can calculate C. By measuring a series of standard solutions, the adsorption quantity of N719 can be obtained. The absorption amount of dye is listed in the Table 3.

Table 3 The electro-chemical characteristics of DSSCs

Electrodes	V oc/mV	J sc/mA cm^−2	FF	η/%	Amount of dye/mol cm^−2 × 10^−7
P25	0.82	8.94	73.07	5.36	1.57
TiO2	0.79	9.57	70.77	5.35	1.15
N/TiO2	0.82	11.10	72.89	6.64	2.26
Double-N/TiO2	0.83	12.40	73.62	7.58	2.52

---

4. Conclusions

In this paper, we investigate the difference between single- and double-N doped TiO₂, and conclude that compared with single-N doped, double-N doped TiO₂ improves open circuit voltage and short circuit current. Furthermore, the double-N doped TiO₂ not only has the advantage of increasing visible light absorption with ammonia as the nitrogen dopant, but also has the advantage of a larger interface area with urea as the nitrogen dopant.¹² Using the double-N doped TiO₂, the photoelectric conversion efficiency has been greatly improved. The double-N doping generated a new microstructure that enhances the specific surface area and porosity, and greatly improves the performance of TiO₂.

Acknowledgements

This work is supported by National Natural Science Foundation of China (no. 21076147), Natural Science Foundation of Tianjin (no. 10JCZDJC23700), National International S&T Cooperation Foundation of China (no. 2012DFG41980) and Independent Innovation Foundation of Tianjin University (2010).

References

1. B. O'Regan and M. Grätzel, Nature, 1991, 353, 737.
2. Q. Wang, S. Ito, M. Grätzel, F. Fabregat-Santiago, I. Mora-Seró, J. Bisquert, T. Bessho and H. Imai, J. Phys. Chem. B, 2006, 110, 25210–25221.
3. U. Bach, et al. , Nature, 1998, 395, 583.
4. M. Grätzel, J. Photochem. Photobiol., C, 2003, 4, 145.
5. J. Zhang, C. Fu, X. Yang and W. Cao, J. Inorg. Organomet. Polym. Mater., 2011, 21, 43–49.
6. J. Im, S. Lee and Y. Lee, Appl. Surf. Sci., 2011, 257, 2164–2169.
7. C. Burda, Y. Lou, X. Chen, A. C. S. Samia, J. Stout and J. L. Cole, Nano Lett., 2003, 3, 1049–1051.
8. S. Yang, H. Xue, H. Wang, H. Kou, J. Wang and G. Zhu, J. Phys. Chem. Solids, 2012, 73, 911–916.
9. R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki and Y. Taga, Science, 2001, 293, 269.
10. R. Nakamura, T. Tanaka and Y. Nakato, J. Phys. Chem. B, 2004, 108, 10617.
11. G. Boschloo and A. Hagfeldt, J. Phys. Chem. B, 2005, 109, 12093.
12. J. Wu, J. Wang, J. Lin, Y. Xiao, G. Yue, M. Huang, Z. Lan, Y. Huang, L. Fen, S. Yin and T. Sato, Sci. Rep., 2013, 3, 2058.
13. W. Guo, Y. Shen, L. Wu, Y. Gao and T. Ma, J. Phys. Chem. C, 2011, 115, 21494.
14. W. Guo, Y. Shen, G. Boschloo, A. Hagfldt and T. Ma, Electrochim. Acta, 2011, 56, 4611–4617.
15. S. Livraghi, M. C. Paganini, E. Giamello, A. Selloni, C. D. Valentin and G. Pacchioni, J. Am. Chem. Soc., 2006, 128, 15666–15671.
16. N. Serpone, J. Phys. Chem. B, 2006, 110, 24287–24293.
17. T. Ma, M. Akiyama, E. Abe and I. Imai, Nano Lett., 2005, 5, 2543–2547.
18. L. Hu, J. Wang, J. Zhang, Q. Zhang and Z. Liu, RSC Adv., 2014, 4, 420–427.
19. N. C. Saha and H. G. Tompkins, J. Appl. Phys., 1992, 7, 3072–3079.
20. L. Qi, C. Li and Y. Chen, Chem. Phys. Lett., 2012, 539–540, 128–132.
21. E. Ramasamy and J. Lee, Energy Environ. Sci., 2011, 4, 2529.
22. X. Qiu and C. Burda, Chem. Phys., 2007, 339, 1.
23. O. Diwald, T. L. Thompshon, T. Zubkov, E. G. Goralski, S. D. Walck and J. T. Yates, Jr, J. Phys. Chem. B, 2004, 108, 6004–6008.
24. K. Selvam and M. Swaminathan, RSC Adv., 2012, 2, 2848–2855.
25. Z. He, Z. Zhu, J. Li, J. Zhou and N. Wei, J. Hazard. Mater., 2011, 190, 133–139.
26. S. Sato, R. Nakamura and S. Abe, Appl. Catal., A, 2005, 284, 131–137.
27. H. Tian, L. Hu, C. Zhang, W. Liu, Y. Huang, L. Mo, L. Guo, J. Sheng and S. Dai, J. Phys. Chem. C, 2010, 114, 1627.
28. Z. Tian, L. Wang, L. Jia, Q. Li, Q. Song, S. Su and H. Yang, RSC Adv., 2013, 3, 6369–6376.
29. S. G. Kim, M. J. Ju, I. T. Choi, W. S. Choi, H. Choi, J. Baek and H. K. Kim, RSC Adv., 2013, 3, 16380–16386.
30. X. Peng, Y. Feng, S. Meng and B. Zhang, Electrochim. Acta, 2014, 115, 255–262.
31. E. Ramasamy and J. Lee, Energy Environ. Sci., 2011, 4, 2529.
32. S. Sadhuab and P. Poddarabc, RSC Adv., 2013, 3, 10363–10369.

---