After preliminary tests in CVD, SEM and EDX analysis indicated that the range of temperature of 550–600 °C has to be avoided for the application of the ITO-substrates as electrodes, since the ITO layer undergoes severe disruption at such high temperatures, becoming no longer conductive. The growth time was also inspected, determining an optimal CNT synthesis time of 30 min. Successful growth of MWCNTs was obtained on Samples A, B and C, treated in CVD for 30 min at 550, 525 and 500 °C respectively ().
SEM images of MWCNTs grown on ITO-coated glass by CVD at: (a) 550 °C, (b) 525 °C, (c) 500 °C.
The transmittance and the resistivity of each electrode were measured and are reported in . The density distribution of the tubes is found to scale considerably with the deposition temperature. On Sample A (550 °C), a dense and thick layer of MWCNTs covers the entire ITO area. Due to the CNT density, the substrate looks almost black and is therefore no longer transparent. In addition, the sheet resistance of the ITO layer increased to 40 Ω/sq (almost three times higher than that for pristine ITO). When heated in air at 550 °C, the polycrystalline ITO layer is known to degrade, leading to the segregation of Sn into clusters hundreds of nanometers in diameter [24
], and also to increased inter-diffusion between the substrate and film [25
]. Both effects can reduce the film conductivity by up to 50%, as reported elsewhere [26
Growth parameters and properties for the three CNT-enhanced electrodes compared to pure glass and to ITO/glass sample.
In our case, the interaction of the ITO film with the CVD process gases at 550 °C is expected to deteriorate the conductivity of the electrode even more strongly. This is partly supported by the formation of microspheres of indium on the ITO film, as observed by SEM and EDX (not shown). Similarly to what was reported by Lan et al. [27
], we suggest that the exposure of the ITO film to a hydrogen atmosphere at 550 °C (and the probable creation of atomic hydrogen coming from the dissociation of either H2
, perhaps enabled by the metal catalyst layer) enables the formation of small clusters of metallic indium, which coalesce during the CVD to form spherical particles with a typical size of >2 µm. As a consequence, the film surface would segregate and change its chemical ratio. The film conductivity will in turn significantly decrease, as will the optical transmittance, on account of a stronger light absorption and scattering caused by those metallic microspheres.
In contrast, at 525 °C (Sample B) and 500 °C (Sample C), the degradation is not as severe and the conductivity of the film is still acceptable (25–30 Ω/sq). In these two cases the nanotubes nucleate with a lower density, and the substrates show a transmittance at 515 nm of 45 and 75%, respectively. illustrates the optical transmittance of these two samples in the wavelength range of 350–750 nm, taking also into consideration the absorption spectrum of the P3HT:PCBM blend.
Transmittance spectra of the electrodes (left Y axis), compared to the absorption spectrum of the P3HT:PCBM blend (right Y axis).
Although on Sample B the density of the CNT carpet is much higher than on Sample C, for the present purpose, an optimal condition is reached with the latter sample. When a temperature of 500 °C is used, the short sparse tubes (average density of 10 tubes/µm2) that grow do not form bundles or thick aggregates, allowing more light to pass through the electrode and to reach the active layer of the cell. SEM images taken in various sites of Sample C (as the one in ) were analyzed to calculate the average dimensions of the grown MWCNTs. The average length of the tubes is 100 nm and the diameter 40 nm, as confirmed by TEM analysis (). Due to the low synthesis temperature the tube structure is very defective and residual allotropes of carbon, such as diamond-like and amorphous carbon, are found around the nanotube walls (confirmed also by Raman spectroscopy, not shown).
(a) SEM image showing the surface of Sample C, on which a low density mat of MWCNTs is grown after CVD at 500 °C for 30 min. (b) TEM image of CNTs from Sample B (grown in CVD for 30 min at 525 °C).
In our context, the presence of defects in the tubular structure could be an advantage in terms of conductivity, because it can induce cross linking between the inner shells (walls) of the tubes through sp3
bond formation, facilitating charge-carrier hopping to inner shells [28
]. Such intershell bridging provides additional charge-carrier transport pathways, offsetting the effect of a conductivity decrease induced by defect scattering.
By measuring various areas of the sample, a mean distribution of 10 MWCNT/µm2
has been estimated, being that the average dimensions of nanotubes are 100 nm in length, 30 nm in diameter, and 58 m2
/g for the specific surface area [29
]. Such values would entail an increase of 10% in the overall surface area of the CNT/ITO electrode in comparison with the planar ITO film. We believe that such three-dimensional and nanostructured electrodes, made of metallic nanotubes [30
], will be able to penetrate the P3HT:PCBM blend and ease the extraction of holes to the external circuit.
Using Sample C, we measured the WF of the as-created electrode. Kelvin probe and ultraviolet photoelectron spectroscopy (UPS) are the techniques usually employed for this purpose; however, there are substantial differences in how the WF is measured. The Kelvin probe method measures, in air, the difference in WF between a millimetric probe and the sample, which can undergo surface reactions with species adsorbed from the environment. Conversely, UPS measures, in ultrahigh vacuum, the lowest WF of a small portion of the surface, usually a few microns in diameter. WF values measured by the Kelvin probe method are often higher than those measured by UPS [31
], due to the influence of the ambient gases and to the fact that the probe size typically covers an area of few millimeters squared. Therefore, we chose to use the Kelvin probe method as it is able to measure the electrode WF in its working environment, just before the cell is built.
After a fine calibration with a reference tantalum foil, the WF of an untreated and clean ITO substrate was found to be 4.80 eV. We then measured a value of 4.95 eV in the case of our CNT-enhanced electrode, that is, an increase of 0.15 eV. Although this value is in good agreement with the WF of MWCNTs reported by Shiraishi et al. [32
], we have to make two considerations: (i) our substrate is not fully covered by a continuous, dense mat of nanotubes; (ii) when measuring by Kelvin probe method, the electrode under test is the whole structure CNT/ITO, not only the CNT overlayer; (iii) a thin layer of Fe is also present between the ITO and the CNT layer, even if during the CVD it should become segregated into small particles, giving rise to the nucleation of the tubes.
All of these occurrences, instead of the sole CNT contribution, would partake in establishing the WF measured for the ITO–CNT electrode (as depicted in ). Nevertheless, this increase in WF is strongly beneficial because it brings the electrode WF closer to that of the photoactive blend. Thus we anticipate a reduction in the hole–injection barrier at the anode interface, as a result of the highest occupied states of ITO–CNT lying lower than those of ITO.
WF levels for cells with ITO (left) and ITO–CNT (right) electrode. (All reported values are in eV and negative).
A similar kind of band alignment is almost achieved in the standard cell architecture by the insertion of a layer of poly(3,4-ethylene dioxythiophene):(polystyrene sulfonic acid) (PEDOT:PSS). This polymer is used to improve the contact (and reduce the mismatch in energy level) between the ITO and the P3HT, although it is also known to shorten the device lifetime [33
]. Being slightly acidic, the PEDOT:PSS is in fact able to etch the ITO and causes interface instability through indium diffusion into the polymer active layer. In our case instead, we believe that using a mat of MWCNTs as a functional buffer layer for ITO should guarantee an increase in both the charge collection and in the lifetime of the device.
In order to test the last statement, test organic solar cells were built with two of our CNT-enhanced anodes: sample C (whose characterization have been presented and discussed above) for cell C, and sample C1 (treated with the same CVD conditions of sample C but for a shorter time of 15 min instead of 30 min) for cell C1. The I–V curve and the output power generated by the cells made with our electrodes are reported in , in comparison with the data obtained for a reference cell made with a standard ITO-coated glass anode (i.e., without the addition of PEDOT:PSS). The I–V characteristic of a standard ITO/PEDOT:PSS/P3HT:PCBM/Al cell is also reported in , for a full understanding of the experimental results. All the numeric values are reported in , along with the respective PCEs.
Current–voltage characteristic and output power of P3HT:PCBM solar cells: (a) Cell C and cell C1, compared to a reference cell made with bare ITO-coated glass; (b) classic ITO/PEDOT:PSS/P3HT:PCBM/Al cell manufactured in our labs.
Characteristics of the organic solar cells: open-circuit voltage (V
oc), short-circuit current density (J
sc), fill factor (FF) and power conversion efficiency (η).
As a preliminary remark, it has to be pointed out that the overall PCE of the experimental cells suffers from the lack of those beneficial effects that are acknowledged by the inclusion of a PEDOT:PSS layer between the ITO and active blend, particularly an advantageous interface morphology [34
] that enables higher J
and fill factor (FF). On the other hand, the comparison between experimental devices made with pristine and CNT-enhanced ITO–glass demonstrates the substantial improvement that the addition of CNTs affords to the electrical properties of the electrode.
By analyzing the I–V graphs, one readily notices how the two CNT-enhanced electrodes dramatically contribute to an increase of the open-circuit voltage (V
oc) of the cell. Remarkably, in the case of cell C1, V
oc reached 272 mV, which is almost three times higher than the value measured for the reference cell made with bare ITO (~100 mV). Such a consistent improvement in V
oc is due to the optimal alignment of the energy levels between the CNT-modified ITO WF (~4.95 eV) and the P3HT HOMO (~5 eV), on the account of a fostered hole collection at the anode/polymer interface. Besides, by taking into consideration an equivalent-circuit diagram for a bulk heterojunction solar cell (), we highlight that the CNTs could be also responsible of a quenched recombination both at the dissociation sites (e.g., donor/acceptor interfaces) and near the anode (as a result of an increase of the shunt resistor R
sh), with a further positive effect on the V
Equivalent circuit of the ideal organic solar cell.
Moreover, we propose that our electrode could contribute to the reduction of the series resistance R
s of the cell by means of the addition of shorter and direct paths for charge collection, which are on average provided by the MWCNTs (having intrinsically a very high aspect ratio). This helps in overcoming the low mobility of the holes, now able to travel more quickly than in the pure P3HT phase, and implies a corresponding increase in J
sc. In particular, the J
sc is expected to benefit from the numerous percolation paths created by the CNTs, which can effectively drive away the free carriers generated from the dissociation of the excitons at the dispersed heterojunctions.
We observe, however, that the J
sc exhibits a noteworthy 40% increase in the case of cell C, but it does not vary much for sample C1. This different behavior for two electrodes prepared with the same procedure must be explained in terms of the only parameter varied, i.e., the CVD time. Consistently with the widely known CNT growth mechanism, the shorter CVD time used for sample C1 (15 min) leads to a shorter length of the grown CNTs: as a result, we speculate that the occurrence of short circuits between the two electrodes should be less likely in this case. Hence, the V
oc is expected to augment correspondingly, while the charge collection and hence the J
sc are less enhanced by the shorter transport paths.
Conversely, the formation of an extra blocking contact (e.g., for holes at the ITO electrode) can be the reason for the small FF values found, which increases only by ~5% in the case of sample C1. This could be considered in the equivalent circuit with the insertion of a counter diode D2 or by another shunt R
sh that directly connects the two electrodes.
As already stated, the absolute efficiency of our cell is not as relevant in the present work as the comparison with the bare ITO cell is. Even without the good ohmic contact provided by PEDOT:PSS, our devices show major improvement in electric performance. In fact, the overall increase in PCE is still more noticeable when considering the lower transparency (due to the CNTs layer) and the higher resistivity (due to thermal and chemical degradation) of the treated ITO film. Nonetheless, CNT-enhanced electrodes may be used in conjunction with a layer of PEDOT:PSS to further advance the PCE of OSCs; or, once the process would be refined, they could become a suitable replacement for PEDOT:PSS, with the aim of improving the interface morphology without compromising the long-term stability of the cell. To this end, more research should be devoted to obtain a more uniform and ohmic contact between the CNTs and the P3HT. Our method could be further improved by exploring very low CVD temperatures (down to 350 °C), which have been reported as being unexpectedly suitable for CNT synthesis from Fe films [35