Mg Doped CuCrO2 as Efficient Hole Transport Layers for Organic and Perovskite Solar Cells

The electrical and optical properties of the hole transport layer (HTL) are critical for organic and halide perovskite solar cell (OSC and PSC, respectively) performance. In this work, we studied the effect of Mg doping on CuCrO2 (CCO) nanoparticles and their performance as HTLs in OSCs and PSCs. CCO and Mg doped CCO (Mg:CCO) nanoparticles were hydrothermally synthesized. The nanoparticles were characterized by various experimental techniques to study the effect of Mg doping on structural, chemical, morphological, optical, and electronic properties of CCO. We found that Mg doping increases work function and decreases particle size. We demonstrate CCO and Mg:CCO as efficient HTLs in a variety of OSCs, including the first demonstration of a non-fullerene acceptor bulk heterojunction, and CH3NH3PbI3 PSCs. A small improvement of average short-circuit current density with Mg doping was found in all systems.


Nanoparticle Synthesis
Mg:CCO nanoparticles with 0 at%, 5 at%, and 10 at% Mg doping levels were synthesized by a hydrothermal method as reported in previous literature [39]. First, 7.5 mmol Cu(NO 3 ) 2 ·2.5H 2 O and stoichiometric amounts of Cr(NO 3 ) 3 ·9H 2 O and Mg(NO 3 ) 2 were dissolved in 35 mL deionized (DI) water and stirred for 15 min at room temperature. Next, 2.5 g NaOH was added into the mixture and stirred for another 15 min at room temperature. The precursor solution was transferred into a 50 mL autoclave reactor (Col-Int Tech., Irmo, SC, USA), filled to 70% of its total volume. The hydrothermal reaction was carried out at 240 • C for CCO and 230 • C for Mg:CCO for 60 hours. Finally, the precipitate was washed using 2 M HCl and EtOH in sequence several times until the supernatant was colorless. After centrifuging, the mud was dried in a desiccator at room temperature overnight to obtain CCO or Mg:CCO powders.

Materials Characterizaton
The crystalline phases of the nanoparticles were characterized by XRD using a Rigaku Ultima III diffractometer (The Woodlands, TX, USA) with Cu Kα (λ = 1.5418 Å) radiation. Powder diffraction files (PDFs) were used to identify characteristic peaks in the XRD patterns. Polytype compositions, crystal size and lattice parameters of XRD patterns were performed using Profex (an open source XRD and Rietveld refinement software, Solothurn, Switzerland) and structure files for phase identification were downloaded from the Crystallography Open Database (COD). The film morphologies were examined using SEM. The experimental Mg doping concentration was quantified using EDX. Nanoparticle size was measured from TEM images from the Delong LVEM5 Benchtop Electron Microscope (Delong TEM, Montreal, QC, Canada) equipped with the Q-Capture Pro 7 software. Hydrodynamic sizes were measured by DLS using a Malvern Zetasizer Nano ZS instrument (Malvern, United Kingdom). The lattice fringe distances were determined from high resolution TEM (HR TEM) images of nanoparticles obtained using a JEOL JEM2100 TEM (Peabody, MA, USA). The elemental compositions and chemical states of the films were analyzed by XPS, using a PHI 5000 Versa Probe II equipped with an Al Kα source and a hemispherical analyzer. XPS data were taken at a 45 • takeoff angle with a pass energy of 23.5 eV. Optical transmission of the films was characterized by UV-vis over the wavelength range from 178 to 890 nm. The band gap energy was determined from Tauc plots of the UV-vis absorbance data. Film thickness was obtained by ellipsometry at 55 • , 65 • , and 75 • incident angles over the wavelength range from 280 to 1690 nm. Ionization energy was measured from 4.7 to 5.8 eV with a 0.05 eV energy step using PESA with deuterium lamp intensity at 100 nW. The work function was measured using a KP apparatus (SKP5050, KP Technology) referenced to Au at 5.15 eV.  61 BM solution on a spinning substrate at 1200 rpm for 60 s, followed by annealing at 170 • C in N 2 for 10 min. Finally, 7 nm Ca and 100 nm Al were sequentially evaporated on top of the active layer. The current-voltage (J-V) measurements were carried out using a 2635A Keithley low-noise sourcemeter under AM 1.5G 100 mW cm −2 illumination from a class AAA solar simulator (Abet Technologies) in a nitrogen filled glovebox. The diode area is 0.11 cm 2 and the aperture area is 0.049 cm 2 .

PFBT2Se2Th:PC 71 BM OSCs
PFBT2Se2Th:PC 71 BM devices were fabricated and tested similarly to the description in Section 2.3.1 unless otherwise noted. Six mg mL −1 PFBT2Se2Th and 12 mg mL −1 PC 71 BM were dissolved in DCB with 5 vol % DPE and stirred at 100 • C overnight. This solution and ITO/CCO or Mg:CCO substrates were preheated at 100 • C. The PFBT2Se2Th:PC 71 BM active layer (~120 nm thick) was made by first dispensing 50 µL of PFBT2Se2Th:PC 71 BM solution on the substrate and then immediately starting spinning at 1200 rpm for 60 s, followed by drying in a vacuum chamber for 2 min.

PTB7-Th:ITIC OSCs
PTB7-Th:ITIC devices were fabricated and tested similarly to the description in Section 2.3.1 unless otherwise noted. PTB7-Th and ITIC were blended in a 1:1 weight ratio, dissolved in a mixed solution (CB with 3 vol% CF) at a total concentration of 20 mg mL −1 and stirred at room temperature overnight. The PTB7-Th:ITIC active layer (~80 nm thick) was made by first dispensing 40 µL PTB7-Th:ITIC solution on the substrate and then immediately starting spinning at 1250 rpm for 60 seconds.

MAPbI 3 PSCs
MAPbI 3 PSCs were fabricated by spin coating the MAPbI 3 layer (~450 nm thick) according to an antisolvent-washing recipe first described by Ahn et al. [42], and thereafter thermally evaporating a C 60 /BCP electron transport layer and then Ag electrodes. The details of the deposition of each of these layers are exactly as described in our previous report [27], except that the antisolvent wash during MAPbI 3 deposition was carried out 11-12 s after starting the spin recipe. After fabrication, the devices were encapsulated with Ossila E131 UV-cure epoxy and a glass coverslip. J.-V curves were measured using a Keithley 2401 sourcemeter and an Oriel solar simulator. The lamp intensity was initially calibrated to 1 sun using a reference Si solar cell from Newport Corp., and maintained at that intensity during the measurements by a reference photodiode. The aperture area of the PSCs is 0.1 cm 2 , as defined by a shadow mask. (J,V) points were collected by sampling the current 1 s after the bias was applied. AC external quantum efficiency measurements were performed using an Enlitech QE-R instrument equipped with a monochromated Xe lamp optically chopped at 165 Hz, and without applied electrical bias. XPS, using a Kratos Analytical Axis Ultra spectrometer equipped with a monochromated Al Kα source, was performed on ITO/HTL/MAPbI 3 films to determine whether diffusion of Cu, Cr, or Mg from the HTL resulted in detectable levels of these elements at the surface of the perovskite film. TRPL experiments comparing ITO/MAPbI 3 and ITO/HTLs/MAPbI 3 structures were performed using a microscope-based time-resolved system [43]. Samples were excited by 405 nm/120 fs optical pulses at 7.6 MHz repetition rate produced by doubling the fundamental frequency of the Mira 900 laser and followed by pulse-picking (1 out of 10 pulses) via the acousto-optical modulator (NEOS Technologies). Excitation of 1 µW was focused on the sample via 0.6 NA objective, which also ensured a high photon collection efficiency to obtain PL signals. The collected emission was passed through a spectrometer and directed either to a CCD camera for PL spectral analysis or to a sensitive photon detector (MicroPhoton Devices MPD 50) for the wavelength-dependent PL lifetime measurements. PL decay curves were collected via the time-correlated single photon counting performed on board of Pico300E photon counting hardware (PicoQuant GmbH). The overall time resolution was better than 200 ps.  Table 1). Rietveld refinement was carried out in order to quantitatively determine the polytype composition, crystal size, and lattice parameters for each XRD pattern. Figure 1b shows the experimental (blue solid circle), calculated (red curve), and difference between experimental and calculated (grey curve) patterns for CCO, 5% Mg:CCO, and 10% Mg:CCO. Table 1 shows the results extracted from the Rietveld refinement. The polytype compositions for all three compounds are found to be~60 ± 3% for 3R and~40 ± 3% for 2H. The crystal sizes decrease monotonically from CCO to 10% Mg:CCO, from 7.8 to 4.5 nm for the 2H polytype calculated from the (004) reflection and from 9.6 to 8.7 nm for the 3R polytype calculated from the (110) reflection. However, the crystal size increases monotonically from 10.2 nm for CCO to 13.1 nm for 10% Mg:CCO in the (110) reflection for the 2H polytype. The tradeoff of the size changes for both polytypes leads to similar widths of the Nanomaterials 2019, 9, 1311 6 of 21 (110) reflection independent of Mg doping. Since delafossite nanocrystals often exhibit anisotropic morphology [25,27], it is reasonable that size changes differ for the (004) and (110) reflections. Similar orientation dependent size variation in Mg:CCO was reported by Bywalez et al. [35]. They attributed the decrease of crystal sizes along the c axis to Mg 2+ obstructing growth of the delafossite crystal structure and stabilizing the spinel phase. However, there is no indication that this phase exists in our samples. Nanomaterials 2019, 9, x FOR PEER REVIEW 6 of 19 orientation dependent size variation in Mg:CCO was reported by Bywalez et al. [35]. They attributed the decrease of crystal sizes along the c axis to Mg 2+ obstructing growth of the delafossite crystal structure and stabilizing the spinel phase. However, there is no indication that this phase exists in our samples. The lattice parameters for 3R-CCO and 2H-CCO polytypes were similar for CCO and 5% Mg:CCO. However, for 10% Mg:CCO, a larger a lattice parameter in both phases (3.00 Å versus 2.99 Å ) and an increase in the c lattice parameter in the 2H-CCO phase (11.46 Å versus 11.43 Å) were observed. This lattice expansion was consistent with Mg substituting on the Cr site, rather than the Cu site [36], because the ionic radius of Mg 2+ (0.72 Å ) is larger than that of Cr 3+ (0.62 Å ) and smaller than that of Cu + (0.77 Å ). This result is consistent with the bond length increase between Cr and O sites after Mg doping predicted from theoretical calculations [33].     1 13.3 13.7 13.4 R exp (%) 2 9.0 9.0 9.1 R p (%) 3 9.0 10. 1 R wp is the weighted profile R-factor and the squared R wp is equal to the weighted sum of squared difference between the experimental and calculated intensity values over the weighted sum of squared experimental intensity values [44]. 2 R exp is the expected R-factor and the squared R exp is equal to the number of data points over the weighted sum of squared experimental intensity values [44]. 3 R p is the profile R-factor and is equal to the weighted sum of difference between the experimental and calculated intensity values over the weighted sum of experimental intensity values [45]. 4 and 5 a and c are the in-plane and out-of-lane lattice constants in the unit cell.

Structural, Compositional, and Morphological Characterizations
The lattice parameters for 3R-CCO and 2H-CCO polytypes were similar for CCO and 5% Mg:CCO. However, for 10% Mg:CCO, a larger a lattice parameter in both phases (3.00 Å versus 2.99 Å) and an increase in the c lattice parameter in the 2H-CCO phase (11.46 Å versus 11.43 Å) were observed. This lattice expansion was consistent with Mg substituting on the Cr site, rather than the Cu site [36], because the ionic radius of Mg 2+ (0.72 Å) is larger than that of Cr 3+ (0.62 Å) and smaller than that of Cu + (0.77 Å). This result is consistent with the bond length increase between Cr and O sites after Mg doping predicted from theoretical calculations [33].
In order to measure Mg concentration in CCO, EDX was performed on 5% Mg:CCO and 10% Mg:CCO. The insets of Figure 2a,b show that Mg is present and distributed uniformly in the Mg:CCO films. Mg/(Mg+Cr) represents the Mg concentration in the Mg:CCO films, which is calculated by atomic number effects (Z), absorption (A), and fluorescence (F) method from EDX spectra in Figure 2 [46]. Table 2 shows that the averaged Mg concentration in 5% Mg:CCO and 10% Mg:CCO measured from EDX is 4.0% and 9.8%, respectively. A possible Mg doping process is proposed similarly to the CCO formation mechanism as described by Miclau et al. [47]. During the hydrothermal synthesis of Mg:CCO nanoparticles, Cu 1+ , Cr 3+ , and Mg 2+ ions can form Cu(OH) − 2 , Cr(OH) − 4 , and Mg(OH) 2 , respectively, at alkaline pH environment according to equations (1-3) below. Mg:CCO nanoparticles can then be formed from the metal hydroxides according to Equation (4) [48][49][50].The formation process of Mg:CCO nanoparticles are given in the following equations: In order to measure Mg concentration in CCO, EDX was performed on 5% Mg:CCO and 10% Mg:CCO. The insets of Figure 2a-b show that Mg is present and distributed uniformly in the Mg:CCO films. Mg/(Mg+Cr) represents the Mg concentration in the Mg:CCO films, which is calculated by atomic number effects (Z), absorption (A), and fluorescence (F) method from EDX spectra in Figure  2 [46]. Table 2 shows that the averaged Mg concentration in 5% Mg:CCO and 10% Mg:CCO measured from EDX is 4.0% and 9.8%, respectively. A possible Mg doping process is proposed similarly to the CCO formation mechanism as described by Miclau et al. [47]. During the hydrothermal synthesis of Mg:CCO nanoparticles, Cu 1+ , Cr 3+ , and Mg 2+ ions can form Cu(OH) 2 -, Cr(OH) 4 -, and Mg(OH)2, respectively, at alkaline pH environment according to equations (1-3) below. Mg:CCO nanoparticles can then be formed from the metal hydroxides according to Equation (4) [48][49][50].The formation process of Mg:CCO nanoparticles are given in the following equations:    Figure 3c) show the nanoparticles exist in individual or small clusters as well as large agglomerates. We only use individual or double nanoparticles (indicated by white circles) to determine particle sizes. The average nanoparticle size for CCO, 5% Mg:CCO, and 10% Mg:CCO is 10.3 ± 2.1, 8.2 ± 2.1, and 9.8 ± 3.0 nm, respectively. The size trend according to TEM results differs slightly from that of Rietveldrefined XRD data. One difference is that the particle size determined from XRD is analyzed for specific reflection and polytype (Table 1), while TEM images are two-dimensional projections of nanoparticles with random orientation. To examine the TEM size results in details, Figure 4a shows box plots of TEM particle sizes for CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue)   2 10.3 ± 2.1 8.2 ± 2.1 9.8 ± 3.0 1 Mg concentration is averaged over five EDX measurements. 2 Nanoparticle size is calculated from TEM images and mean size for each sample is averaged over 50 individual nanoparticles ( Figure 3).
TEM images of CCO (Figure 3a), 5% Mg:CCO (Figure 3b), and 10% Mg:CCO ( Figure 3c) show the nanoparticles exist in individual or small clusters as well as large agglomerates. We only use individual or double nanoparticles (indicated by white circles) to determine particle sizes. The average nanoparticle size for CCO, 5% Mg:CCO, and 10% Mg:CCO is 10.3 ± 2.1, 8.2 ± 2.1, and 9.8 ± 3.0 nm, respectively. The size trend according to TEM results differs slightly from that of Rietveld-refined XRD data. One difference is that the particle size determined from XRD is analyzed for specific reflection and polytype (Table 1), while TEM images are two-dimensional projections of nanoparticles with random orientation. To examine the TEM size results in details, Figure 4a shows box plots of TEM particle sizes for CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) nanoparticles. The ranges of CCO and 5% Mg:CCO nanoparticle sizes are smaller than that of 10% Mg:CCO nanoparticles. It is clear that 90% of the CCO nanoparticles are larger than 8 nm. In contrast, significant fractions of both types of Mg:CCO nanoparticles are between 6 and 8 nm. Thus, the statistics of TEM results show that Mg:CCO samples have greater numbers of smaller-sized particles, although 10% Mg doping appears to broaden the distribution. Considering the particle size results from XRD and TEM, overall Mg doping decreases CCO nanoparticle sizes because the XRD results show the size decreases along the c axis and TEM results show greater numbers of smaller sized Mg:CCO particles.    There are no significant differences among the three doping concentrations due to possibly similar hydrodynamic layer thickness. This is expected because the hydrodynamic layer is determined by solution ionic strength and hydrodynamic size and is often larger than dry particle size [51,52]. Since the Mg concentration in these nanoparticles is low, it is not surprising that hydrodynamic sizes for all samples are similar.
The HR TEM image of 5% Mg:CCO nanoparticles shows clear lattice fringes ( Figure 5). The lattice spacing of 2.47 Å corresponds to the (012) reflection for 3R-CCO polytype. The lattice spacing of 2.33 Å corresponds to the (102) reflection for 2H-CCO polytype. No other lattice spacings corresponding to impurity phases are detected, consistent with XRD results.     There are no significant differences among the three doping concentrations due to possibly similar hydrodynamic layer thickness. This is expected because the hydrodynamic layer is determined by solution ionic strength and hydrodynamic size and is often larger than dry particle size [51,52]. Since the Mg concentration in these nanoparticles is low, it is not surprising that hydrodynamic sizes for all samples are similar.
The HR TEM image of 5% Mg:CCO nanoparticles shows clear lattice fringes ( Figure 5). The lattice spacing of 2.47 Å corresponds to the (012) reflection for 3R-CCO polytype. The lattice spacing of 2.33 Å corresponds to the (102) reflection for 2H-CCO polytype. No other lattice spacings corresponding to impurity phases are detected, consistent with XRD results.  There are no significant differences among the three doping concentrations due to possibly similar hydrodynamic layer thickness. This is expected because the hydrodynamic layer is determined by solution ionic strength and hydrodynamic size and is often larger than dry particle size [51,52]. Since the Mg concentration in these nanoparticles is low, it is not surprising that hydrodynamic sizes for all samples are similar.
The HR TEM image of 5% Mg:CCO nanoparticles shows clear lattice fringes ( Figure 5). The lattice spacing of 2.47 Å corresponds to the (012) reflection for 3R-CCO polytype. The lattice spacing of 2.33 Å corresponds to the (102) reflection for 2H-CCO polytype. No other lattice spacings corresponding to impurity phases are detected, consistent with XRD results. XPS studies were carried out in order to confirm the oxidation states of Cu, Cr, and Mg in the Mg:CCO powders. XPS data was analyzed using PHI Multipak software and peak fitting was done using a Gaussian-Lorentzian profile after a Shirley type background subtraction [53]. The binding energy was shifted using the valence band edge. The measured (cross symbol) and fitted (solid curve) XPS spectra of Cu 2p3/2, Cr 2p3/2, Mg 1s, and O 1s core levels for 5% Mg:CCO (red color) and 10% Mg:CCO (blue color) nanoparticles are shown in Figure 6. Deconvolution of the Cu 2p3/2 spectrum for both 5% and 10% Mg:CCO (Figure 6a) results in two peaks at 934.6 eV and 932.3 eV corresponding to binding energies of Cu(OH)2 and Cu 1+ , respectively, consistent with the literature [54,55]. The Cr 2p3/2 spectrum (Figure 6b) can be fitted to two peaks at 577.3 eV and 576.5 eV corresponding to binding energies of Cr 3+ as hydroxide and Cr 3+ as oxide, respectively [37,56]. These are similar to the binding energy peak positions of Cu 1+ and Cr 3+ oxide of undoped CCO nanoparticles reported in the literature [24]. Figure 6c shows the Mg 1s spectra, wherein the peak is located at binding energy of 1303.1 eV, corresponding to the Mg 2+ oxidation state [39]. The O 1s spectrum (Figure 6d) shows peaks corresponding to lattice oxygen (OI) at 529.9 eV and hydroxyl groups (OII) at 531.5 eV for both 5% Mg:CCO and 10% Mg:CCO nanoparticles. A small-intensity peak at 533 eV corresponding to adsorbed water for 5% Mg:CCO nanoparticles is observed [54,57]. (c) (d) Figure 6. X-ray photoelectron spectroscopy (XPS) spectra of (a) Cu 2p3/2, (b  XPS studies were carried out in order to confirm the oxidation states of Cu, Cr, and Mg in the Mg:CCO powders. XPS data was analyzed using PHI Multipak software and peak fitting was done using a Gaussian-Lorentzian profile after a Shirley type background subtraction [53]. The binding energy was shifted using the valence band edge. The measured (cross symbol) and fitted (solid curve) XPS spectra of Cu 2p 3/2 , Cr 2p 3/2 , Mg 1s, and O 1s core levels for 5% Mg:CCO (red color) and 10% Mg:CCO (blue color) nanoparticles are shown in Figure 6. Deconvolution of the Cu 2p 3/2 spectrum for both 5% and 10% Mg:CCO (Figure 6a) results in two peaks at 934.6 eV and 932.3 eV corresponding to binding energies of Cu(OH) 2 and Cu 1+ , respectively, consistent with the literature [54,55]. The Cr 2p 3/2 spectrum (Figure 6b) can be fitted to two peaks at 577.3 eV and 576.5 eV corresponding to binding energies of Cr 3+ as hydroxide and Cr 3+ as oxide, respectively [37,56]. These are similar to the binding energy peak positions of Cu 1+ and Cr 3+ oxide of undoped CCO nanoparticles reported in the literature [24]. Figure 6c shows the Mg 1s spectra, wherein the peak is located at binding energy of 1303.1 eV, corresponding to the Mg 2+ oxidation state [39]. The O 1s spectrum (Figure 6d) shows peaks corresponding to lattice oxygen (O I ) at 529.9 eV and hydroxyl groups (O II ) at 531.5 eV for both 5% Mg:CCO and 10% Mg:CCO nanoparticles. A small-intensity peak at 533 eV corresponding to adsorbed water for 5% Mg:CCO nanoparticles is observed [54,57].
binding energy peak positions of Cu 1+ and Cr 3+ oxide of undoped CCO nanoparticles reported in the literature [24]. Figure 6c shows the Mg 1s spectra, wherein the peak is located at binding energy of 1303.1 eV, corresponding to the Mg 2+ oxidation state [39]. The O 1s spectrum (Figure 6d) shows peaks corresponding to lattice oxygen (OI) at 529.9 eV and hydroxyl groups (OII) at 531.5 eV for both 5% Mg:CCO and 10% Mg:CCO nanoparticles. A small-intensity peak at 533 eV corresponding to adsorbed water for 5% Mg:CCO nanoparticles is observed [54,57].

Optical and Electronic Characterizations
The thickness of CCO and Mg:CCO nanoparticle films are controlled by the number of coating cycles that were performed during deposition [27]. Figure 7a-c shows SEM images for CCO, 5% Mg:CCO, and 10% Mg:CCO films; no regions of bare substrate are seen for all films. Figure 8a shows the UV-vis absorbance and transmission (inset) spectra of well-covered CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) films. The absorbance values at 300 nm lie between 0.21 and 0.22 for all films, which are highly transparent (transmission > 90%) in the visible region. All three films are 18 nm thick as determined by ellipsometry. The direct band gap (E g ) is extrapolated from the Tauc plot ( Figure 8b). The average values of the direct E g are 3.27 ± 0.02 eV, 3.25 ± 0.03 eV, and 3.27 ± 0.03 eV for CCO, 5% Mg:CCO, and 10% Mg:CCO, respectively (Table 3). These values are the same within the uncertainty of the measurement (~0.03 eV). Thus, Mg doping does not affect the direct E g of CCO. The E g for pure CCO is consistent with our previous result [27].

Optical and Electronic Characterizations
The thickness of CCO and Mg:CCO nanoparticle films are controlled by the number of coating cycles that were performed during deposition [27]. Figure 7a-c shows SEM images for CCO, 5% Mg:CCO, and 10% Mg:CCO films; no regions of bare substrate are seen for all films. Figure 8a shows the UV-vis absorbance and transmission (inset) spectra of well-covered CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) films. The absorbance values at 300 nm lie between 0.21 and 0.22 for all films, which are highly transparent (transmission > 90%) in the visible region. All three films are 18 nm thick as determined by ellipsometry. The direct band gap (Eg) is extrapolated from the Tauc plot ( Figure 8b). The average values of the direct Eg are 3.27 ± 0.02 eV, 3.25 ± 0.03 eV, and 3.27 ± 0.03 eV for CCO, 5% Mg:CCO, and 10% Mg:CCO, respectively (Table 3). These values are the same within the uncertainty of the measurement (~ 0.03 eV). Thus, Mg doping does not affect the direct Eg of CCO. The Eg for pure CCO is consistent with our previous result [27].        percentile WF value for the 10% Mg:CCO is higher than that of CCO and 5% Mg:CCO. The 5% Mg:CCO films exhibit the largest spread with a long tail to the large WF than CCO films. Thus, Mg:CCO films generally appear to have higher WF values than CCO films, although the difference is below the level of statistical significance. Figure 9b shows box plots of the ionization energy (IE) for CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) films. The median IE values of CCO, 5% Mg:CCO, and 10% Mg:CCO films are 5.11, 5.08, and 5.06 eV, respectively. The energy step for the IE measurement is 0.05 eV. The 50-percentile IE value decreases monotonically with increasing Mg concentration. Furthermore, the measured IE value is consistent with the IE of 5.1 eV from previous band structure calculations [27]. As shown in Figure 9, the overall WF values are larger than IE values, especially for Mg:CCO films. Thus, these films are p-type degenerately doped. The difference between WF and IE values (WF -IE) increases with Mg concentration from 0.08 eV for CCO to 0.16 eV for 10% Mg:CCO (Table 3), indicating that Mg:CCO films may have higher conductivity, consistent with previous results [35,36].
with the IE of 5.1 eV from previous band structure calculations [27]. As shown in Figure 9, the overall WF values are larger than IE values, especially for Mg:CCO films. Thus, these films are p-type degenerately doped. The difference between WF and IE values (WF -IE) increases with Mg concentration from 0.08 eV for CCO to 0.16 eV for 10% Mg:CCO (Table 3), indicating that Mg:CCO films may have higher conductivity, consistent with previous results [35,36].  Figure 10a and Table 4 show the results of P3HT:PC61BM devices with CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. The average Jsc of P3HT:PC61BM devices is higher with Mg:CCO HTL (from 6.94 mA cm -2 for CCO to ~7.05 mA cm -2 for Mg:CCO). The average Voc of Mg:CCO is also higher than that of undoped CCO HTL (~0.582 V versus 0.570 V). However, both Mg:CCO devices exhibit lower average FF (0.642 for 5% doping and 0.666 for 10% doping versus 0.685 for no doping). The tradeoff of the three parameters leads to similar PCE values for all devices independent of Mg doping. We note that the variation among different diodes is larger in Jsc than Voc or FF, which is typical of OPV devices. Nonetheless, there is a systematic trend of average Jsc increase with Mg doping. Figure 10b and Table 4 Figure 10c and Table 4 show the results of PTB7-Th:ITIC devices with CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. The average Jsc of the devices increases monotonically from 11.55 mA cm -2 for undoped CCO HTL to 12.02 mA cm -2 for 10% Mg:CCO HTL. The average Voc and FF are highest for the 5% Mg:CCO in this batch of devices, but they do not depend on Mg doping in other batches. Generally, the PCE values of PTB7-Th:ITIC devices are higher when using Mg:CCO as  Figure 10a and Table 4 show the results of P3HT:PC 61 BM devices with CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. The average J sc of P3HT:PC 61 BM devices is higher with Mg:CCO HTL (from 6.94 mA cm −2 for CCO to~7.05 mA cm −2 for Mg:CCO). The average V oc of Mg:CCO is also higher than that of undoped CCO HTL (~0.582 V versus 0.570 V). However, both Mg:CCO devices exhibit lower average FF (0.642 for 5% doping and 0.666 for 10% doping versus 0.685 for no doping). The tradeoff of the three parameters leads to similar PCE values for all devices independent of Mg doping. We note that the variation among different diodes is larger in J sc than V oc or FF, which is typical of OPV devices. Nonetheless, there is a systematic trend of average J sc increase with Mg doping. Figure 10b and Table 4 show the results of PFBT2Se2Th:PC 71 BM devices with CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. The average V oc and FF are similar in all devices with values of~0.665 V and~0.685, respectively. The average J sc of the devices increases monotonically from 10.50 mA cm −2 for undoped CCO HTL to 10.88 mA cm −2 for 10% Mg:CCO HTL. Thus, the PCE values of PFBT2Se2Th:PC 71 BM devices are higher when Mg:CCO, instead of undoped CCO, is used as the HTL. Among different diodes, the variation of J sc is larger than that of V oc or FF. Moreover, Mg:CCO devices have even larger variation of J sc . However, a similar systematic trend of increasing average J sc as P3HT:PC 61 BM devices is observed in PFBT2Se2Th:PC 71 BM devices. Figure 10c and Table 4 show the results of PTB7-Th:ITIC devices with CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. The average J sc of the devices increases monotonically from 11.55 mA cm −2 for undoped CCO HTL to 12.02 mA cm −2 for 10% Mg:CCO HTL. The average V oc and FF are highest for the 5% Mg:CCO in this batch of devices, but they do not depend on Mg doping in other batches. Generally, the PCE values of PTB7-Th:ITIC devices are higher when using Mg:CCO as the HTL due to the increase in J sc . This work is the first using CCO and Mg:CCO as HTL for BHJ OSCs with a non-fullerene acceptor. Figure 10d and Table 4 show the results of MAPbI 3 PSCs with CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs under forward (solid lines) and reverse scans (dashed lines). Only slight hysteresis is seen, indicating minimal trap states at the CCO or Mg:CCO/perovskite interface. Under forward scan, the average V oc increases monotonically from 0.985 V for the undoped CCO HTL to 1.007 V for the 10% Mg:CCO HTL. Similar trends are observed in the J sc (from 18.91 mA cm −2 for the undoped CCO HTL to 19.40 mA cm −2 for the 10% Mg:CCO HTL) and FF (from 0.678 for the undoped CCO HTL to 0.703 for the 10% Mg:CCO HTL). Overall, the PCE of the devices improves monotonically from 12.64% for the undoped CCO HTL to 13.73% for the 10% Mg:CCO HTL. We note that the variation among different diodes is large for all the parameters. However, there are systematic trends of increases among the average J sc , V oc , and FF with Mg doping. Under reverse scan, the average J sc increases monotonically from 18.70 mA cm −2 for the undoped CCO HTL to 19.37 mA cm −2 for the 10% Mg:CCO HTL. The FF of the devices using the 5% Mg:CCO and 10% Mg:CCO HTLs were similar, 0.719, but for the undoped CCO HTL, a lower FF (0.697 versus 0.719) was observed. The V oc is similar in all devices with values of~1.01 V. Overall, the PCE of the devices improves monotonically from 13.19% for the undoped CCO HTL to 14.12% for the 10% Mg:CCO HTL. As in the forward scan data, despite the variation among different diodes, there is a systematic trend of increasing average J sc with Mg doping. Jeong et al. observed that Mg:CCO produces PSCs with a slightly higher J sc and V oc , but a lower FF, resulting in no improvement in the PCE; however, they did not report Mg concentration [40].

CCO and Mg:CCO as HTLs in OSCs and PSCs
The inset in Figure 8d shows the external quantum efficiency (EQE) at wavelength ranging from 300 to 800 nm for CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. A broadband increase in EQE is observed with Mg doping, consistent with the increases of average J sc in forward and reverse J-V scans.
the undoped CCO HTL to 19.40 mA cm for the 10% Mg:CCO HTL) and FF (from 0.678 for the undoped CCO HTL to 0.703 for the 10% Mg:CCO HTL). Overall, the PCE of the devices improves monotonically from 12.64% for the undoped CCO HTL to 13.73% for the 10% Mg:CCO HTL. We note that the variation among different diodes is large for all the parameters. However, there are systematic trends of increases among the average Jsc, Voc, and FF with Mg doping. Under reverse scan, the average Jsc increases monotonically from 18.70 mA cm -2 for the undoped CCO HTL to 19.37 mA cm -2 for the 10% Mg:CCO HTL. The FF of the devices using the 5% Mg:CCO and 10% Mg:CCO HTLs were similar, 0.719, but for the undoped CCO HTL, a lower FF (0.697 versus 0.719) was observed. The Voc is similar in all devices with values of ~1.01 V. Overall, the PCE of the devices improves monotonically from 13.19% for the undoped CCO HTL to 14.12% for the 10% Mg:CCO HTL. As in the forward scan data, despite the variation among different diodes, there is a systematic trend of increasing average Jsc with Mg doping. Jeong et al. observed that Mg:CCO produces PSCs with a slightly higher Jsc and Voc, but a lower FF, resulting in no improvement in the PCE; however, they did not report Mg concentration [40]. The inset in Figure 8d shows the external quantum efficiency (EQE) at wavelength ranging from 300 to 800 nm for CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. A broadband increase in EQE is observed with Mg doping, consistent with the increases of average Jsc in forward and reverse J-V scans. Figure 10. Average J-V curves (number of devices for each system is given in the footer of Table 4)   Several groups have reported elemental diffusion from inorganic transport layer into MAPbI 3 when using CdS ETL [58], CrO x [59], and CuI [60] HTLs. In order to examine this possibility, we performed XPS studies on the surfaces of MAPbI 3 films on top of ITO/HTL. Figure 11 shows the normalized XPS spectra of (a) survey, (b) Cu 2p, (c) Cr 2p, and (d) Mg 2p core levels for MAPbI 3 films processed on top of CCO (black), 5% Mg:CCO (red), and 10% Mg:CCO (blue) HTLs. For all HTLs, all peaks in the survey spectra are indexed as the component elements (C, N, Pb, and I) of MAPbI 3 , consistent with our previous result [27]. The Cu 2p and Cr 2p spectral ranges are free of any peaks corresponding to Cu 2p or Cr 2p (dashed lines), indicating no presence of Cu and Cr elements at the surface of MAPbI 3 layer. The Mg 2p spectrum shows two peaks at 48.2 eV and 46.6 eV corresponding to the binding energy of the I 4d orbitals [61]. No Mg 2p peak at 50.8 eV (dashed line) was detected, indicating no presence of Mg element. Thus, if metal diffusion from CCO and Mg:CCO HTLs into MAPbI 3 occurs, it does so at a level below the sensitivity of XPS. This result is consistent with thermodynamic calculation: the calculated formation enthalpy of CCO is -6.0 eV [22], significantly lower compared to that of CdS (−1.5 eV) [62] and CuI (−0.3 eV) [63] and slightly lower compared to that of Cr 2 O 3 (−5.9 eV) [64]. Mg:CCO has the same crystalline structure as CCO and the Mg doping content is small, thus, the formation enthalpy of Mg:CCO is expected to be similar to that of CCO. Thus, CCO and Mg:CCO are more stable and less likely to decompose or react than the aforementioned HTLs. Nevertheless, additional experimentation is warranted to explore the possibility of reactivity between CCO/Mg:CCO and perovskite phases. Nanomaterials 2019, 9,   The positions of Cu 2p1/2, Cu 2p3/2, Cr 2p1/2, and Cr 2p3/2 peaks are indexed according to our previous CCO reports [24,27]. The position of Mg 2p peak is indexed according to the report from Hoogewijs et al. [65]. In (d), the peaks correspond to the I 4d orbitals; peaks due to the Mg 2p orbitals are not observed.
In order to explore charge transport at the CCO and Mg:CCO/MAPbI3 interface, we performed TRPL measurements. Figure 12a Figure 12b shows the PL lifetimes extracted from three exponential fits in all samples (lines in Figure 12b). The τ1, τ2, and τ3 lifetimes of the ITO/MAPbI3 structure are 1.5 ns, 4.9 ns, and 16.3 ns, respectively. After adding CCO and Mg:CCO HTLs, the τ1, τ2, and τ3 decreases to 0.7 ns, ~3.0 ns, and ~11.0 ns, respectively, indicating enhanced charge extraction and consistent with the literature result [66]. Again, there are no significant differences in PL lifetimes among films of MAPbI3 on CCO and Mg:CCO HTLs. Mg doping in CCO is expected to lead lower PL intensity and shorter lifetimes. However, these effects are not discernable in our TRPL results, presumably because they may be confounded by factors besides charge transfer, such as surface recombination [67].  [24,27]. The position of Mg 2p peak is indexed according to the report from Hoogewijs et al. [65]. In (d), the peaks correspond to the I 4d orbitals; peaks due to the Mg 2p orbitals are not observed.
In order to explore charge transport at the CCO and Mg:CCO/MAPbI 3 interface, we performed TRPL measurements. Figure 12a shows the PL emission spectrum for ITO/MAPbI 3 (green), ITO/CCO/MAPbI 3 (black), ITO/5% Mg:CCO/MAPbI 3 (red), and ITO/10% Mg:CCO/MAPbI 3 (blue). For all samples, the main PL emission peaks are at~750 nm, consistent with the literature [66]. PL intensities for MAPbI 3 on top of CCO and Mg:CCO HTLs are lower compared to that of MAPbI 3 on ITO, indicating CCO and Mg:CCO HTLs are effective in promoting charge transfer. However, PL intensities are similar among MAPbI 3 on top of CCO and Mg:CCO HTLs. Figure 12b shows the normalized TRPL decay kinetics for ITO/MAPbI 3 (green), ITO/CCO/MAPbI 3 (black), ITO/5% Mg:CCO/MAPbI 3 (red), and ITO/10% Mg:CCO/MAPbI 3 (blue). With the addition of CCO and Mg:CCO HTLs, a faster PL decay is observed relative to ITO/MAPbI 3 . The inset table in Figure 12b shows the PL lifetimes extracted from three exponential fits in all samples (lines in Figure 12b). The τ 1 , τ 2 , and τ 3 lifetimes of the ITO/MAPbI 3 structure are 1.5 ns, 4.9 ns, and 16.3 ns, respectively. After adding CCO and Mg:CCO HTLs, the τ 1 , τ 2 , and τ 3 decreases to 0.7 ns,~3.0 ns, and~11.0 ns, respectively, indicating enhanced charge extraction and consistent with the literature result [66]. Again, there are no significant differences in PL lifetimes among films of MAPbI 3 on CCO and Mg:CCO HTLs. Mg doping in CCO is expected to lead lower PL intensity and shorter lifetimes. However, these effects are not discernable in our TRPL results, presumably because they may be confounded by factors besides charge transfer, such as surface recombination [67].   PSCs are higher with Mg:CCO HTLs. The small average Jsc increases in all systems may be partially attributed to the better conductivity of Mg:CCO HTLs resulting from the increased WF with respect to IE with Mg doping. Additionally, the broadband increase with Mg doping content in the EQE spectra of PSCs (Figure 10d inset) signifies that the increased HTL work function contributes to a stronger electric field within the device, more efficiently extracting photoexcited carriers regardless of the depth at which the generating photons are absorbed. If Voc and FF are independent of Mg doping, the PCE may be expected to increase due to the boost in Jsc. However, they do not show a consistent trend from batch to batch. Voc and FF are more susceptible to film roughness, which can vary due to aggregation of the nanoparticles in the suspensions and variation in spin coating conditions. The tradeoff between Jsc and Voc/FF results in little or no statistical PCE improvement (Table 4).    PSCs are higher with Mg:CCO HTLs. The small average Jsc increases in all systems may be partially attributed to the better conductivity of Mg:CCO HTLs resulting from the increased WF with respect to IE with Mg doping. Additionally, the broadband increase with Mg doping content in the EQE spectra of PSCs (Figure 10d inset) signifies that the increased HTL work function contributes to a stronger electric field within the device, more efficiently extracting photoexcited carriers regardless of the depth at which the generating photons are absorbed. If Voc and FF are independent of Mg doping, the PCE may be expected to increase due to the boost in Jsc. However, they do not show a consistent trend from batch to batch. Voc and FF are more susceptible to film roughness, which can vary due to aggregation of the nanoparticles in the suspensions and variation in spin coating conditions. The tradeoff between Jsc and Voc/FF results in little or no statistical PCE improvement (Table 4).  In (a,b), the applied bias is at 0.8 V. In (c), the applied bias is initially at 0.85 V. After 50 s, it switches to 0.8 V. Figure 14 shows the bar charts of average J sc for P3HT:PC 61 BM OSCs, PFBT2Se2Th:PC 71 BM OSCs, PTB7-Th:ITIC OSCs, and MAPbI 3 PSCs under forward and reverse scans for CCO (black color), 5% Mg:CCO (red color), and 10% Mg:CCO (blue color) HTLs. The average J sc of all OSCs and MAPbI 3 PSCs are higher with Mg:CCO HTLs. The small average J sc increases in all systems may be partially attributed to the better conductivity of Mg:CCO HTLs resulting from the increased WF with respect to IE with Mg doping. Additionally, the broadband increase with Mg doping content in the EQE spectra of PSCs (Figure 10d inset) signifies that the increased HTL work function contributes to a stronger electric field within the device, more efficiently extracting photoexcited carriers regardless of the depth at which the generating photons are absorbed. If V oc and FF are independent of Mg doping, the PCE may be expected to increase due to the boost in J sc . However, they do not show a consistent trend from batch to batch. V oc and FF are more susceptible to film roughness, which can vary due to aggregation of the nanoparticles in the suspensions and variation in spin coating conditions. The tradeoff between J sc and V oc /FF results in little or no statistical PCE improvement (Table 4).

Conclusions
In summary, we synthesized CCO and Mg:CCO nanoparticles and successfully applied them as HTLs in OSCs and PSCs. Mg incorporation induces a slight lattice expansion by substituting larger ionic radii Mg 2+ into the Cr 3+ site. Rietveld refinement suggests that Mg doping decreases CCO nanoparticle size along the c axis but increases CCO nanoparticle size along the in-plane directions. Overall, both XRD and TEM results indicate that nanoparticle sizes are smaller with Mg doping. The average value of the direct Eg is (3.26 ± 0.03) eV in all nanoparticle films. The WF values for all Mg concentrations are larger than the IE values, and their difference (WF -IE) increases with Mg concentration, consistent with increased p-type conductivity reported in the literature. OSCs and PSCs based on Mg:CCO HTLs show a consistent increase in average Jsc in all four absorber systems despite large uncertainties; however, an overall enhancement in PCE is not clearly discernible (except in PSCs) due to different trends in other parameters and sample variation. No elemental (Cu, Cr, and Mg) diffusion from CCO and Mg:CCO HTLs is detected by XPS at the surface of MAPbI3 films. CCO and Mg:CCO HTLs effectively extract charge from the absorber, as evident in more PL quenching and shorter lifetimes when MAPbI3 is deposited on the HTLs. Mg doping in CCO HTLs enhances the stabilized efficiency for MAPbI3 PSCs. This work provides new insights related to the role that an Mg:CCO HTL may play in improving performance in a wide range of OSCs and MAPbI3 PSCs.

Conclusions
In summary, we synthesized CCO and Mg:CCO nanoparticles and successfully applied them as HTLs in OSCs and PSCs. Mg incorporation induces a slight lattice expansion by substituting larger ionic radii Mg 2+ into the Cr 3+ site. Rietveld refinement suggests that Mg doping decreases CCO nanoparticle size along the c axis but increases CCO nanoparticle size along the in-plane directions. Overall, both XRD and TEM results indicate that nanoparticle sizes are smaller with Mg doping. The average value of the direct E g is (3.26 ± 0.03) eV in all nanoparticle films. The WF values for all Mg concentrations are larger than the IE values, and their difference (WF -IE) increases with Mg concentration, consistent with increased p-type conductivity reported in the literature. OSCs and PSCs based on Mg:CCO HTLs show a consistent increase in average J sc in all four absorber systems despite large uncertainties; however, an overall enhancement in PCE is not clearly discernible (except in PSCs) due to different trends in other parameters and sample variation. No elemental (Cu, Cr, and Mg) diffusion from CCO and Mg:CCO HTLs is detected by XPS at the surface of MAPbI 3 films. CCO and Mg:CCO HTLs effectively extract charge from the absorber, as evident in more PL quenching and shorter lifetimes when MAPbI 3 is deposited on the HTLs. Mg doping in CCO HTLs enhances the stabilized efficiency for MAPbI 3 PSCs. This work provides new insights related to the role that an