Understanding the Role of Surface States on Mesoporous NiO Films

Surface states of mesoporous NiO semiconductor films have particular properties differing from the bulk and are able to dramatically influence the interfacial electron transfer and adsorption of chemical species. To achieve a better performance of NiO-based p-type dye-sensitized solar cells (p-DSCs), the function of the surface states has to be understood. In this paper, we applied a modified atomic layer deposition procedure that is able to passivate 72% of the surface states on NiO by depositing a monolayer of Al2O3. This provides us with representative control samples to study the functions of the surface states on NiO films. A main conclusion is that surface states, rather than the bulk, are mainly responsible for the conductivity in mesoporous NiO films. Furthermore, surface states significantly affect dye regeneration (with I–/I3– as redox couple) and hole transport in NiO-based p-DSCs. A new dye regeneration mechanism is proposed in which electrons are transferred from reduced dye molecules to intra-bandgap states, and then to I3– species. The intra-bandgap states here act as catalysts to assist I3– reduction. A more complete mechanism is suggested to understand the particular hole transport behavior in p-DSCs, in which the hole transport time is independent of light intensity. This is ascribed to the percolation hole hopping on the surface states. When the concentration of surface states was significantly reduced, the light-independent charge transport behavior in pristine NiO-based p-DSCs transformed into having an exponential dependence on light intensity, similar to that observed in TiO2-based n-type DSCs. These conclusions on the function of surface states provide new insight into the electronic properties of mesoporous NiO films.


S2
precursor is better to be kept inside the chamber for a longer time before purging directly.
In our previous work, we tried to ALD TiO2 or ZnO (including Al2O3) inside mesoporous NiO. But we found TiO2 or ZnO more likely localized on the top of the NiO films, which caused that formation of an electronically connected TiO2 or ZnO layer inside NiO film was problematic [3] . That may be due to that an excess of precursors at each cycle was difficult to be purged away from the NiO. The accumulated precursors reacted with next precursor and blocked the pore on the top of the NiO layer. This could also due to that precursor diffusion was more difficult through the windingly routes in NiO. As a result, the deposition did not reach a monolayer adsorption at each cycle.
Therefore, typical ALD procedures of planer film should be reconsidered when it is applied on a mesoporous film. More attentions should be taken especially only several of ALD cycles. One cycle of ALD should be more careful when applied on a mesoporous films since the precursor adsorption is more easily affected. This can result in diverse situation only due to the possible adsorption of the chamber. Not to say that a mesoporous film over 10 μm in thickness is used.

Films Color:
Figure S1. The images of NiO film as prepared (left) and NiO-A film after ALD of Al2O3 (right).

UV-vis and Near IR Transmittance Spectroscopy:
Transmittance spectroscopy of the dry film NiO and NiO-A was performed at a Lambda 900 double-beam UV/vis/NIR spectrophotometer (PerkinElmer) with an integrating sphere, and a Spectralon reflectance standard. The scatter and reflection light were carefully subtracted.

Crystallization Structure of the NiO and NiO-A Films:
X-ray Diffraction (XRD) was performed in a Siemens D5000 θ−2θ goniometer with CuKα (λ = 1.54051 Å) radiation. XRD measurement with a small angle X-ray irradiation was used to achieve the XRD spectra for the thin films.  Figure S3 XRD spectra of NiO (black) and NiO-A (red) dry film. The labeled peaks belong to NiO or NiO-A, and the non-labeled peaks belong to FTO glasses.

The Density of States (DOS) of NiO and NiO-
where p is porosity, e is the elementary charge, and l is the film thickness. p is assumed as 0.5 in this paper [4] .  If there were no intra bandgap states in both NiO and NiO-A films, extracting same amount of charges will be corresponding to an identical Voc. Voc will be equal to the difference potential of Quasi-Fermi level in NiO or NiO-A and the redox potential (treated as a fixed value). In the Figure S5, the value of Voc was significantly different at a fixed charge extracted. The Quasi-

S5
Fermi level in NiO-A was positively shifted 270 mV compared to that in NiO due to the passivation of the intra bandgap states.

Direct Current (DC) Conductivity Measurement:
In the four-probe resistance measurement, a semiconductor device parameter analyzer (Keysight B1500A, Keysight Technologies, Inc.) was adopted to force the current and measure the voltages with the four source units (SMU, B1517A, Keysight Technologies, Inc.). In order to avoid the stochastic error, the forced current was swept from zero to a certain value (according to the sample resistance). At each current level, the potentials at the two inner electrodes were recorded. Afterwards, the resistance can be extracted by linear fitting the whole potential-current curves. All the electrical measurements were taken in a Faraday cage with the electrical connection through a probe station.
The illumination condition was created by using a white LED lamb (PAR38 module, 17 W, 5000K, Zenaro Lighting GmbH, 420-750 nm) located above the samples ca. 10 cm which is corresponding to a simulated one sun condition at the target wavelength range. The strips of FTO on the glass were made by etching away those original FTO layers from the white area labeled as 'Glass'. Before etching, the SCOTCH tape in 5 mm width was tapped on the strips of FTO as shown Scheme S1. Then, the rest part of the exposed FTO was cast a thin layer of Zn metal powders. Then, the aqueous solution of 2 M HCl was added on the Zn powders to etch away FTO layer. After etching, the resistance on the glass area should be carefully checked to make sure all the FTO layers have been removed.
Scheme S1. The NiO or NiO-A samples for resistance measurements with dimension information; and the sketched description of the four-probe resistance measurement.  :  Table S2. The performance of p-DSC based on NiO and NiO-A film.

Parameters of Solar Cells
Footnotes: All the solar cell were measured at stimulated AM 1.5 G, 100 mW·cm -2 light condition, and active area was 0.25 cm 2 (a black mask of 5*5 mm). The electrolyte is 0.05 M I2, 0.1 M LiI in acetonitrile. a The peak IPCE value is added.   Figure S7, the onset potential of the photocurrent is -S7 0.02 V and 0.07 V for NiO and NiO-A film, receptively (by assuming the onset potential is at the position where photocurrent density is 0 mA/cm 2 ). The dark and the light current basically keep constant from -0.2 V to -0.8 V. A clear difference in Figure S7 is that dark and photocurrent density from NiO photocathode are significant larger (ca. 24 times for dark current density, and ca. 20 times for photocurrent density) that of NiO-A photocathode. The experiments clearly show the different photoelectrochemical responses of NiO and NiO-A under dark or light conditions.

CV Measurement with Different Working Electrodes:
The CV measurements were performed inside a glove box (H2O < 0.1 ppm; O2 < 0.1 ppm) with S8 dry ACN solvent and dry supporting electrolyte TBAPF6. Form the CV of TiO2 in I -/I3 -, we did not see the reduction peak around -0.5 V in Figure S12. It is can be explained that there is no DOS in this potential range since there should be inside the bandgap range from -1.0 V to 2.5 V (vs. Ag/AgNO3; the edge of CB in TiO2 is -1.0 V). However, there are significant DOS from -1.0 V to -0.5 V (trap states in TiO2). Still, no reduction peak can be observed, which should support our conclusion that TiO2 cannot catalyze I3reduction.  Figure S14. CV of NiO and NiO-A * (of some bench of samples) as working electrode collected in 5 mM I -/I3 -(5mM I2 and 10 mM LiI) and 100 mM TBAPF6 in ACN at 100 mV/s. NiO-A in Figure 5 is added for comparison.
In some bench of samples, we found the surface states cannot be removed in the range of valence band, named as NiO-A * , see Figure S13. The Re2 peak in such sample can be observed at -0.08 V, see Figure S14, is corresponding to (Re2). Combined with the CV in NiO-A in Figure  5, removing parts of the surface states in VB leads to completely loss the Rx2. It probably suggests that those surface states in the valence band range should directly relate to the active sites of catalysis in Re2.

Uv-vis Absorption of NiO-PB6 and NiO-A-PB6
were measured at Varian Cary 50. Basically, similar absorption is seen from NiO-PB6 and NiO-A-PB6 films. Light harvested efficiency S12 (LHE) also is calculated by the equation LHE=1-10 -A (A is adsorption). It shows ca. 70 % LHE (at the peak) can be achieved on both of NiO-PB6 and NiO-A-PB6 films.   From the DAS, the signature of reduced PB6 (PB6 -) can be observed ca. 650 nm in both NiO-PB6 and NiO-A-PB6 films, matched well with previous reported work. We did not directly observe absorption shoulder of PB6from TAS in both NiO-PB6 and NiO-A-PB6 as previous study [5] . The reason is that absorption spectra of excited PB6 (PB6*) and PB6are mixed together. However, the black lines in DAS of NiO-PB6 and NiO-A-PB6 films can be reasonably assigned as the component of excited PB6. Both DAS of PB6* show same decay rate constant (τ = 1.6 ps), which means the kinetics of hole injection from PB6*into VB of NiO are identical in both NiO-PB6 and NiO-A-PB6 films. Namely, mono/sub-monolayer of ALD of Al2O3 does not affect hole injection, and ALD of Al2O3 are basically transparent to electrons (hole injection efficiency can be estimated to unity).
Fitting Details of Figure 6 in Main Text.
The global fitting of multiple wavelengths was performed in software of Glotaran with a triple exponential function after deconvoluting from a Gaussian shaped instrument response function (IRF). The fitted kinetic traces in Figure 6 were from the same global fitting model. The corresponding kinetic parameters were inserted in DAS of Figure S19 (NiO-PB6: τ1 = 1.6 ps, τ1 = 21 ps, and τ3 = 2.4 ns; NiO-A-PB6: τ1 = 1.6 ps, τ1 = 39 ps, and τ3 = 9.9 ns), and the amplitude components can be compared at targeted wavelength (650 nm) in Figure S19. The fitting residual in Figure 6 was given below. From hole lifetime (τHL) and transport time (τHT), the charge collection efficiency (ηcc) is calculated with equation ηcc =1-(τHT/τHL). Beyond 70% charge collection efficiency (under one sun condition) is obtained on p-DSCs based on NiO-A film.
As known, IPCE can be described as The measured IPCE of NiO-A based solar cells is < 1 %. From calculation above, it is reasonable to treat the dye regeneration (<0.02 %; far different from the calculated dye regeneration efficiency in NiObased p-DSCs [3] ) as the performance limiting step in NiO-A based solar cells. Figure 6).  Figure S21. The function of transport time on light intensity (the current density shows linear dependence on light intensity) from NiO and NiO-A based p-DSCs. (exactly same experiment is repeated as Figure 6 and transport time in this Figure S21 is cited from our previous paper [3] ).

Conductivity Measurement on nonporous NiO Film:
A planar NiO film with thickness of 140 nm was prepared by sputter. The sheet resistance (Four Point Probe, model CMT-SR2000N) is 2.5*10 5 Ω/sq, conductivity of 0.29 S/cm. Then, a modified one cycle of ALD Al2O3 was performed to check the resistance change (named as p-NiO and p-NiO-A).
The clear importance of surface states on NiO film should relate to the characteristic of nanostructure. If the surface states have the tendency to localize on the surface compared to bulk, making it into nanostructure will magnify such effect to many orders of magnitude, due to the larger ratio of surface to volume on mesoporous films. To prove this, a planer NiO film (p-NiO) was prepared by sputtering and corresponding p-NiO-A film by modified one cycle of ALD Al2O3. The DC dark conductivity of p-NiO and p-NiO-A is compared, but no conductivity change is found after ALD (both are 0.39 S/cm). This means that there are different conductivity properties of planer and mesoporous NiO films, resulting from the particular nanostructure. The conductivity of single crystal NiO with stoichiometric composition is proved to have significant a high electrical resistance. One theory claims the conductivity of NiO (without perfect stoichiometric composition in reality) results from the presence of Ni 3+ in the lattice and those Ni 3+ /holes are tending to delocalize in the lattice which causes the conductivity [6] . With this theory, the Ni 3+ could be more likely distribute on the surface on NiO, making the conductivity difference on the mesoporous NiO film.

Estimation of the number of surface states per area.
The surface area of NiO film is estimated about 70 m 2 /cm 3 . The concentration of the surface states in NiO and NiO-A is about 2.78*10 20 cm -3 and 8.05*10 19 cm -3 , respectively. Thus, 4 surface states is in 1 nm 2 of NiO surface area, and 1 surface state is in 1 nm 2 for NiO-A. In contrast, the trap states in TiO2 mesoporous film can also be estimated about 1 in 1 nm 2 according to the ref [7] . It supports that the more surface states exist in NiO mesoporous films. Binding energy/ eV Figure S22. Ni 2p3/2 spectrum of p-NiO.  XPS experiments were conducted on a PHI Quantera II scanning XPS microprobe (Chanhassen, NM). The spectra were calibrated against the C 1s peak at 284.8 eV for adventitious carbon. S17 Table S3. Details parameters of peak A and peak B in Ni 2p3/2 XPS spectra shown in Figure  S23. To further understand the properties of the surface states on NiO, XPS was performed. The underlying chemical nature of surface state of NiO has been investigated by many groups [8,9] , and it is commonly characterized by XPS due to its sensitivity to the surface. However, the reported results are controversial and ambiguous, especially for the indication of Ni oxidation states in Ni 2p3/2 XPS spectrum. Here, we do not try to assign a XPS peak to a corresponding Ni oxidation state. XPS measurements are applied to prove that the typical surface feature of NiO is still kept in NiO-A, as confirmed in Figure S23 [the overall Ni 2p3/2 XPS spectra are changed in which peak A and B is merged together in the spectrum of NiO-A, and the area ratio of B:A is very different (1.6 for NiO, 0.63 for NiO-A)]. In the XPS measurements, Ar ions with relatively low energy are commonly used to clean the surface of samples. It has been also reported that metal oxides could be reduced by the Ar ions if the energy is significantly high [10] . Interestingly, from our sputter-cleaning measurements with Ar ions (Figure S23 middle), we are able to observe a clear Ni metal peak (around 852.5 eV) formed in NiO, but not in NiO-A (see discussion below). This suggests that the surface states on NiO are easier to be reduced than that on NiO-A.

XPS Measurement for NiO and NiO-A Films.
Form XPS studies in NiO film, we know the peak A is belong to Ni 2+ state. However, the explanation for the shoulder of peak B is controversial. The peaks B is known really sensitive to the change of surface states. Here, we do not try to assign peak B to any Ni oxidation states. A confident conclusion here is the surface states in NiO is more reductively active than the surface states in NiO-A.

Photoelectrochemical Properties of Dye sensitized NiO and NiO-A in acetate buffers
with different pH. LSV measurements are similar as Figure S7 except that an Ag/AgCl reference electrode (in 3M NaCl) was used.  NiO-A-PB6 in acetate buffer Current density/ mA/cm