Nanostructured photovoltaics

Figure 1(b) also shows loss mechanisms that DSSCs are susceptible to. These include direct recombination of the electron and hole pairs in the dye, and recombination of the electrons in the ETL with the oxidized dye or the electrolyte [17, 18]. Hence, optimizing charge injection from the dye to the ETL is crucial for efficient device operation. The dye molecule is responsible for harvesting the incident light energy, resulting in a long-lived excited state, followed by electron injection into the conduction band of the ETL. Two different charge transfer pathways have been identified via femtosecond transient absorption spectroscopy: electron injection from non-relaxed higher-lying excited state, which has been associated with the singlet excited state of the dye, and injection from the fully relaxed excited state, which is associated to the triplet excited state of the dye [18, 19]. These processes compete with back electron transfer from the ETL to the dye, implying that for efficient power conversion, fast electron transfer and slow back electron transfer are needed.

The electron kinetics have been studied extensively in the Ru(dcbpy)₂(NCS)₂ (RuN3)–TiO₂ system, as it is one of the most efficient dye/ETL combinations. Since the lowest excited state of RuN3 is 0.3 eV higher than the conduction band minimum of TiO₂, electron injection can occur from the singlet and triplet states of RuN3 [19–22]. By comparing transient absorption signals of the dye deposited on TiO₂ and the dye in solution, a shorter time constant was observed for the former case, indicating that a depopulation channel is available. Furthermore, by probing the triplet state absorption, it was concluded that excited state evolution from the singlet to the triplet state is also occurring, with a time constant of approximately 70 fs [19]. The fast component of electron injection from the singlet states has been estimated to occur on a time scale of <100 fs [19, 23–26]. On the other hand, the slow component of electron injection, associated with the transfer from the triplet states of the dye, has been associated with time constants ranging from 1.7 to 100 ps [19, 24–27]. These processes are depicted in figure 2.

Zoom In Zoom Out Reset image size

Figure 2 also shows the energy dependence of the states in the conduction band of TiO₂, which has been associated with the variation of the electron injection rate, as it was observed to increase with higher density of states. This has been used to rationalize the presence of a slow component of electron transfer, as the triplet states are close to the energy of the conduction band minimum of TiO₂, meaning that there is a reduced density of states available [19, 22, 28]. In addition, while the fast component of electron transfer could occur adiabatically, the slow component could be associated with non-adiabatic processes resulting from the formation of the triplet state [19]. The slow component could also be a consequence of the binding of the dye on the ETL and surface defects [19]. On the other hand, back electron transfer has been observed to happen non-exponentially and is associated with timescales of microseconds to milliseconds, and a negligible picosecond/nanosecond component, therefore reducing competition for the aforementioned electron transfer processes [18, 19]. This difference in timescales justifies the high efficiency of the RuN3–TiO₂ system in DSSCs. The slow back electron transfer has been attributed to weak electronic coupling between the oxidized dye and the electron in the ETL, trapping of the electron in the ETL, as well as the inverted Marcus region that slows outer sphere electron transfer [19, 29].

Electron injection studies have been extended to other dyes as well. Work on the Ruthenium bipyridyl sensitizer dye N719 has demonstrated that unlike the RuN3 dye, in which electron injection from the singlet and triplet states is occurring in parallel, electron injection from the triplet state of the N719 dye is the more efficient injection pathway [30]. It was estimated that singlet injection occurs within 1–10 ps, triplet injection occurs within ∼100 ps, whereas intersystem crossing from the singlet to the triplet state occurs within ∼100 fs [19] and decay of the triplet state to ground state requires ∼10 ns. Therefore, Koops et al [30] concluded that electron injection from singlet states occurs only for systems with very favorable interfacial energetics, as is the case in the RuN3–TiO₂ system, but for typical devices, due to the presence of the electrolyte which influences the potential, it is unlikely that singlet injection can compete with intersystem crossing. Consequently, electron injection will most likely originate from the triplet state, which results in a ten-fold retardation of injection. This work [30] also investigated the various factors that could influence the electron injection kinetics in a DSSC. While the application of electrical bias or the mere presence of the electrolyte in the device did not affect the injection kinetics, it was found that the composition of the electrolyte had a larger effect. The concentration of tert-butylpyridine (tBP) and lithium cations (Li⁺) significantly modified kinetics. Incorporation of tBP in the electrolyte is observed to shift the density of states of TiO₂ to more negative potentials, whereas Li⁺ results in less negative potentials. The former changes the conduction band of TiO₂ by coordination to the TiO₂ surface, or by diminishing the cation concentration on the surface [30–32]. On the other hand, Li⁺ adsorb or intercalate into the TiO₂ layer, eventually modifying the charge and energy levels [28, 30, 33]. However, it was demonstrated that despite the retardation of injection kinetics because of these electrolyte additives, resulting in reduction of the device photocurrent, their presence increases the Fermi level of TiO₂ and, consequently, enhances the open circuit voltage of the device, therefore resulting in higher device efficiency. This further establishes the fact that it is not necessary to have the faster injection kinetics for best device performance, but rather a good compromise between driving force that allows competition with dye relaxation and raising the conduction band of TiO₂ to reduce recombination losses [30]. Hence, the only requirement is that electron injection from the dye to the ETL happens faster than decay of the excited state.

Although Ruthenium bipyridyl dyes were originally believed to be the best choice for DSSCs, record devices have been prepared with porphyrin sensitizers [16, 34]. Therefore, it is imperative that we discuss the electron injection kinetics of these sensitizers and how they compare with Ruthenium bipyridyl dyes. Ruthenium bipyridyl dyes are characterized by long-lived excited state, good visible absorption, and good photochemical stability [21], as well as a high degree of charge transfer character and mixed singlet/triplet states as discussed previously. On the other hand, other common dyes, including porphyrins, do not show as much mixing of singlet/triplet states and charge transfer, but they also demonstrate long-lived singlet excited states [26, 35–37]. The photophysical characteristics of the two types of dyes are different when in solution; RuN3 exhibits very low emission yield and large red shift of emission, associated with the relaxation of the singlet state to the triplet states, whereas porphyrin dyes (zinc and free base tetracarboxyphenyl porphyrins: ZnTCPP and H₂TCPP respectively) show a series of narrow visible absorption bands and small red shift with high emission intensity, characterized by relatively long excited state lifetime [26]. After deposition of all dyes on TiO₂, almost indistinguishable, multi-exponential injection kinetics are observed for all, with essentially the same lifetimes and relative amplitudes. This result demonstrates that while for porphyrin dyes electron injection occurs from the singlet state, and for ruthenium bipyridyl dyes charge transfer and relaxation to the triplet state is involved, it does not dominate electron injection kinetics. Furthermore, the recombination kinetics for RuN3 and ZnTCPP are also indistinguishable, whereas H₂TCPP demonstrates recombination kinetics that are eight times slower. This result further supports the model that recombination in dyes is mostly controlled by the electron transfer between trap sites in the mesoporous ETL, rather than the electronic structure of the dye [26]. Once the electron is injected into the ETL mesoporous layer, it is transported to the contact, so it can be extracted from the device. Materials such as ZnO and SnO₂ have been used in DSSCs, but nanocrystalline porous TiO₂ appears to be the best option so far, due to good energy level matching with dyes [38, 39]. The efficient electron transport through the nanocrystalline TiO₂ layer is believed to occur by electron hopping from one particle to another [18, 20]. Mesoporous layers are unlike the bulk material as they are characterized by low conductivity, the particles making up the layer are too small to support an electrical field, whereas the mesoporous layer allows for formation of interpenetrating networks that result in a large contact area with the other components of the device [18].

The transport of electrons through the ETL is a complex procedure that has been investigated by multiple research groups. Since the electrons are transported through a trapping and de-trapping mechanism, the traps play a crucial role for both electron transfer and recombination in TiO₂ [23, 40, 41]. Traps become important at low illumination intensity, resulting in slow transport and a small diffusion coefficient. At higher light intensity, deep traps are filled, meaning that electron transport is faster and the diffusion coefficient is larger [18, 42]. An additional mechanism regarding the charge transport of TiO₂ films has been suggested, indicating that self-doping is occurring under illumination, which results in enhanced conductivity of the originally poorly conductive material [38]. Under illumination, a transient charging process is happening, and when carrier concentration of 10¹⁹ cm⁻³ is reached, the region of the film that is not illuminated undergoes the Mott (insulator-metal) transition, resulting in a significant increase in electrical conductivity [38, 43]. Modification of the TiO₂ layer has been attempted by various groups to improve its characteristics, including doping of TiO₂ with Zn, Al, Zr, Nb [39, 44, 45], or preparation of 1D nanostructures to enhance electron transport [46].

For efficient DSSCs, electrolytes must be chosen such that they can transport the charge carriers between anode and cathode successfully, considering the redox potential, regeneration of the dye and the electrolyte. Furthermore, it must ensure good contact with the mesoporous ETL, have minimal leakage and evaporation, as well as long-term stability. In addition, it should not result in dye degradation or desorption, and should not absorb a significant portion of visible light [47–50]. In a DSSC, the electrolyte is acting as a hole-transport medium, and is responsible for regenerating the oxidized dye, with the cycle completed after conversion of the triiodide to iodide at the counter electrode. Therefore, the counter electrode must be chosen to ensure low overpotential and rapid reaction [17, 47, 51]. Platinum (Pt) has been a good candidate for a counter electrode. In liquid electrolytes, ion diffusion occurs via ion hopping or liquid-like diffusion of ionic sub-lattice [47, 52]. However, when there is a high concentration of iodide, polyiodides are formed, resulting in an electron exchange mechanism to be responsible for carrier transportation [47, 53, 54]. In this case, electrons are transported via chemical bond exchanges [53, 55]. Overall, the diffusion process is controlled by the size of the redox species, the solvent viscosity, the redox medium concentration, and the separation between electrodes [47, 56].

In the first example of DSSC, a liquid electrolyte prepared with the iodide/triiodide couple in an organic solvent was used [14]. Liquid electrolytes offer several attractive characteristics: they can be prepared easily, they are highly conductive, they have low viscosity, and can achieve good interfacial contact with the counter electrode [47, 48, 56]. They remain the most common choice for hole transport, and were used in the most efficient DSSCs so far [16, 57]. The liquid electrolyte usually comprises of the solvent, the ionic conductor, and the additives. Good solvents must have melting points below −20 °C, and boiling points above 100 °C to prevent evaporation during cell operation, they must be chemically stable within the operating potential window of the device, have low viscosity to ensure that the redox species has high diffusion coefficient and high conductivity, it should not absorb incident light, it should not react with the sensitizer dye, not dissolve the sealant material, be non-toxic and cheap [47]. While polar organic solvents and ionic liquids are good candidates, a single solvent cannot satisfy all the conditions, therefore solvents are usually mixed. Water was initially used as a solvent in DSSCs, but the oxidation of iodide to iodate, which is not reduced by the counter electrode, prevented long-term operation of the device, whereas many dyes are sensitive to water [17, 47, 56, 58]. Acetonitrile is another good electrolyte as it has low viscosity, good chemical stability and solubility, and was used in the DSSC that achieved record efficiency of 13% [16, 47]. Nevertheless, it suffers from low boiling point and high toxicity, meaning it cannot be implemented in industrial devices. Methoxyacetonitrile and 3-methoxypropionitrile have been used in DSSCs, with the latter demonstrating particularly good stability, retaining 98% of the efficiency after 1000 h of accelerated heat tests under thermal stress [17, 47, 59]. Other common solvents include ethylene carbonate, propylene carbonate, γ-butyrolactone, and N-methyloxazolidinone, which are characterized by high boiling and melting points [47]. A DSSC prepared with γ-butyrolactone solvent has been shown to operate outdoors for almost 2.5 years [60]. It should be noted that basic solvents affect the TiO₂ surface, raising its flatband potential, and subsequently not only reducing the driving force for electron injection from the dye to the TiO₂, but additionally preventing electron recombination between the TiO₂ and the redox species [15, 47, 61]. This results in lower device photocurrent, but higher open circuit voltage.

Ionic liquids have also been incorporated in the electrolytes in DSSCs, as they possess good chemical and thermal stability and high ionic conductivity, while their viscosity can be tailored and they do not evaporate or leak as much [47]. Methyl-hexyl-imidazolium iodide was used in a device that demonstrated impressive stability in 1996 [53]. Later, an electrolyte made with SeCN⁻/${{\left({\rm{SeCN}}\right)}_{3}}^{-}$ and without any solvent was implemented, achieving high efficiency at the time (7.5%–8.3%) [62]. The good performance was attributed to the low viscosity, high conductivity, and low light absorption of the ionic liquid, but the stability of the device was not good. Though imidazolium-based ionic liquids are common, other materials that have been used in electrolytes include ionic liquids with sulfonium, guanidinium, ammonium, pyridinium, and phosphonium, but none of these has shown good efficiencies [63–68]. However, implementation of a tetrahydrothiophenium melt demonstrated that ionic liquids can also be used for high efficiency devices [69]. Pure ionic liquids are characterized by high viscosity and low ion mobility, which results in lower diffusion coefficient of the triiodide, in comparison to organic solvents [70]. Hence, in order to improve their performance, ionic liquids are often mixed with organic solvents, which produced devices of high efficiency [71]. Ionic liquids have also been mixed with solid components to improve their conductivity [72]. Mixing with multi-walled carbon nanotubes resulted in significant improvement of device efficiency [73].

The iodide/triiodide redox couple has traditionally been used in DSSCs for dye regeneration, as it has a redox potential that matches that of the dye, thereby ensuring rapid dye regeneration and slow recombination. Furthermore, this redox couple is soluble, conductive, does not absorb as much visible light, has good stability, and can form good interfaces with the mesoporous ETL [47]. One of the issues of DSSC operation is the recombination of charge carriers at the interface between the ETL and the electrolyte, which can be suppressed by incorporation of a metal oxide blocking layer [74]. It can also be suppressed by introducing additives such as 4-tert-butylpyridine and guanidium thiocynate in the electrolyte [21, 47, 75–77]. Similarly, dyes that have hydrophobic chains also appear to prevent recombination by blocking the contact of the two materials [78], whereas other dyes have the reverse effect of increasing the amount of triiodide near the dye [79, 80]. In order to increase the rate at which the dye is regenerated, the concentration of iodide must be tailored depending on whether organic solvent or ionic liquid is used; higher concentration is needed for the latter due to its lower viscosity [81]. Even though the iodide/triiodide redox species has been used for a large number of DSSCs, it suffers from issues of encapsulation due to the high vapor pressure of iodide and corrosion of the sealant by the iodide, while the polyiodides absorb part of the incident light, limiting the photocurrent [47]. In addition, the mismatch between the redox potential of the dye and that of the redox species causes a reduction of the possible open circuit voltage that can be achieved [47, 82]. As a result, alternative complexes were pursued. These include bromide/tribromide [51], interhalogen redox systems [83], disulfide/thiolate [84], and most notably, cobalt complexes [16, 34, 57]. Cobalt complexes have been used in both devices that currently hold the efficiency record [16, 57] and are considered to be the most efficient redox species for DSSCs, with further improvements expected after tuning the redox potentials of the redox couple [47]. Other metal complexes that have been studied include Ni(III)/Ni(IV), Cu(I)/Cu(II), and ferrocene/ferrocenium [85–89].

Electrical additives are often included in the electrolyte to optimize device performance. As mentioned in sections 2.1.1, 4-tert-butylpyridine (tBP) is a common additive, which improves the open circuit voltage, by shifting the conduction band edge of TiO₂ to more negative values [47]. Additives with similar effects include various nitrogen-containing compounds, such as pyridine, alkylaminopyridine, alkylpyridine, benzimidazole, pyrzaole, quinolone, among others [47]. Another type of additive, also mentioned in the previous section, are lithium ions (Li⁺) or guanidinium ions, which adsorb to the surface of TiO₂ and assist in electron injection from the dye and reduce electron recombination by attaching to the surface of the TiO₂ layer, thereby improving the device photocurrent. Both types of additives are often combined in devices in order to acquire benefits from each one: raising the conduction band of TiO₂ and minimizing recombination losses [90–92].

Liquid electrolytes suffer from instability and leakage issues under exposure to air and prolonged storage. Therefore, research focus shifted to developing solid-state HTMs to avoid such problems. The redox electrolytes discussed in section 2.1.2 can transport charge carriers via ion movement and can be essentially considered a hole-transport layer. Thus, they could be replaced by p-type semiconductors, which are also capable of hole-transport via hole hopping through neighboring molecules [47]. A good HTM must have its valence band maximum above the ground state energy level of the dye to allow for efficient hole transfer and should be characterized by good hole mobility [93–95]. Furthermore, it should not absorb visible light, or cause dye degradation. [96, 97] In addition, as sufficient pore filling of the mesoporous ETL is crucial, the HTM should allow for deposition in an amorphous state to ensure good contact between the various layers of the DSSC [98, 99]. Some of the first types of inorganic solid-state HTMs included copper compounds (CuI, CuBr, CuSCN) [100–103]. These materials allow for solution or vacuum deposition, and are characterized by good hole conductivity [97]. CuI was incorporated in a DSSC for the first time in 1995, with low efficiency, which was attributed to the rapid crystallization of CuI that prevented sufficient filling of the mesoporous ETL and hence, caused insufficient electrical contact [95, 100]. After addition of 1-methyl-3-ethylimidazolium thiocyanate (MEISCN), which acts as a growth inhibitor for CuI, improved pore filling was achieved, leading to increased device efficiency of 3.8% [104]. Nevertheless, devices based on CuI HTM decay rapidly, because of the formation of Cu₂O and CuO caused by the iodine molecules on the surface of CuI acting as hole trapping sites [47, 105–107]. Introduction of a MgO blocking layer increased device efficiency by preventing hole transfer from TiO₂ to CuI [108]. The low efficiency of these devices could also be attributed to the insufficient contact between the dye and the HTM. However, it was observed that dyes with thiocyanate ligands can be strongly bound to CuI films, improving contact and therefore, device efficiency [109]. Further modification of the Cul HTM and the counter electrode with guanidine thiocyanate enhanced device performance, reaching device efficiency of 7.4% [110, 111].

CuSCN was investigated as an alternative to CuI, as it could offer higher stability [47, 101]. However, CuSCN suffers from low hole conductivity which necessitates material engineering to improve its electrical characteristics. Doping with (SCN)₂ led to the formation of acceptor levels below the CuSCN band gap, whereas inclusion of Cu(II) sites in CuSCN increased its hole conductivity, with both modifications resulting in improved device performance [112, 113]. Another promising solid-state HTM is cesium tiniodide (CsSnI₃), which demonstrates high hole conductivity and hole mobility, and a band gap of 1.3 eV [47]. Since CsSnI₃ can be prepared by solution processing, it allows for good pore filling of the mesoporous ETL, whereas its low band gap improves light absorption at longer wavelengths. Its incorporation in a DSSC resulted in device efficiency of 3.72% and doping of CsSnI₃ with SnF₂ raised the efficiency to 6.81% [47]. Through additional device engineering, CsSnI₃-based DSSCs reached efficiency of 10.2% [114]. Further modifications include use of Cs₂SnI₆ instead of CsSnI₃, which is stable in air and moisture, allowing for device fabrication in air [115].

Organic HTMs offer certain benefits over inorganic materials, such as abundancy, low cost, and facile preparation [47, 116]. and can be broadly classified into polymeric and molecular HTMs [116]. Polymeric HTM polypyrrole was first implemented in a DSSC in 1997, but it performed poorly due to absorption of visible light [117–119]. Polyaniline was also introduced as a potential HTM, but only reached device efficiency of 1.15% [120, 121]. Poly(3-hexylthiophene) (P3HT) and poly(3-octylthiophene) (P3OT) have dominated the polymeric HTM research landscape since then. While initial device efficiencies were low due to insufficient pore filling of the mesoporous TiO₂, device engineering boosted efficiency to 1.3% and demonstrated good stability after 3-month storage in air [122–125]. Device efficiencies were further enhanced by addition of bis(trifluoromethanesulfonyl)imide (LiTFSI) and tBP to P3HT reaching 2.7% [126, 127], as well as successful infiltration of P3HT into TiO₂ nanotubes [128]. Combination of P3HT with [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) further increased device efficiency to 4.11% [129]. One of the most promising polymeric HTMs is poly(3,4-ethylenedioxythiophene) (PEDOT), which is characterized by minimal absorption in the visible part of the spectrum, good hole conductivity, and good stability at room temperature [47]. DSSCs incorporating PEDOT have reached efficiency of 6.8% [130]. Among molecular HTMs, 2,2',7,7'-tetrakis(N,N-di-p-methoxyphenylamine)-9,9'-spirobifluorene (spiro-OMeTAD) is one that has shown remarkable results [47]. It initially resulted in device efficiency of 0.74% due to charge recombination at the spiro-OMeTAD/ETL interface [93], but through doping of the material with FK102 cobalt(III) complex that increased its hole mobility, and combination with Y123 sensitizer, device efficiency of 7.2% was achieved [131]. Even though solid state HTMs show better long-term stability, they still lag in device efficiency in comparison to their liquid-based counterparts, due to electrical contact issues with the rest of the device layers [47].

While significant improvements have been made in the materials incorporated in DSSCs, there are still issues to be resolved before the efficiency can be enhanced further. Extensive work has been done on understand and addressing each of these.

2.1.4.1. Absorption in the near-infrared spectral region

The first record efficiencies in DSSC research were achieved by implementation of the ruthenium N3, N719, and black N749 dyes. These dyes performed well as sensitizers in the visible part of the spectrum, but failed to absorb well in the near-infrared part of the spectrum, which, as shown in figure 3, is a significant portion of the solar spectrum [17, 132]. Therefore, it is imperative to develop dyes with improved response in the region of the spectrum beyond 900 nm to achieve high power conversion efficiencies. To meet this requirement, various types of sensitizers for DSSCs have been developed. Metal complexes, including ruthenium, have been popular due to their broad absorption spectrum. These sensitizers consist of a metal ion and ancillary ligands with an anchoring group, with the absorption in the visible part of the spectrum associated with a metal to ligand charge transfer [17]. The ligands are tuned to modify the properties of the sensitizer to achieve higher device performance.

Zoom In Zoom Out Reset image size

Ruthenium N3 dye has been commonly used due to its broad absorption in the visible part of the spectrum, leading to efficiencies of 10% [15]. By modifying the ligands to extend the absorption spectrum to the near-infrared region, the N749 'black dye' was designed, which successfully extended IPCE to 920 nm, and resulted in PCE of 10.4% [133, 134]. Other efforts include preparation of a ruthenium dye with a tetradentate ligand, which extended the absorption spectrum to 900 nm, resulting in efficiency of 5.9% [135]. Due to a modification of the HOMO and LUMO levels of the dye, improved spectral response was achieved, resulting in device efficiency of 8.7% [136]. Introduction of thiophene ligands was observed to alter the energy levels of the dye, red-shifting the absorption spectrum [17]. Therefore, this lead to a 10 nm red shift of the absorption, and a device efficiency of 9.5%, when the modified N719 was implemented [137]. Other metal complexes were also investigated in the efforts to achieve improved near-infrared absorption. By replacing the ruthenium ion with osmium, the IPCE spectrum was extended to 1100 nm, while maintaining good device performance [138].

Porphyrins are also of significant interest in the effort to achieve good absorption in the near-infrared part of the spectrum. They demonstrate strong absorption in the Soret band (400–450 nm) and the Q-band (500–700 nm) [139], indicating that they can operate as panchromatic sensitizers, with efforts made to enhance the Q-band. Porphyrin dyes with a donor-π-acceptor structure, with porphyrin as the donor, and cyanoacrylic acid as the acceptor, can be tuned through various thiophene derivatives which act as the π bridge and can be used to enhance the absorption of the dye [17, 140]. Such a dye was used to achieve record device efficiency of 13% with a single sensitizer, as addition of the proquinoidal benzothiadiazole resulted in broadening of the Soret and Q-band absorption, reaching 800 nm [16]. Phthalocyanines are another type of dye that demonstrate high absorption around 700 nm, and are thus candidates for good absorption in the near-infrared part of the spectrum [17]. Nevertheless, these dyes suffer from poor solubility and aggregation on the ETL surface, necessitating the use of a co-adsorber [17]. A Zn-phthalocyanine dye has demonstrated maximal IPCE in the near-infrared region, but the device efficiency was only 0.57% [141].

Extensive work has also been done on this front on organic dyes, as they also possess a donor-π-acceptor structure, which makes tuning of their absorption spectra easy [17]. Vinylene units were introduced into coumarin sensitizers to extend their absorption spectrum, producing dyes characterized by broad red-shifted absorption, due to a shift of the HOMO level to higher energy, and resulting in device efficiency of 6% [17, 142]. Device performance based on coumarin dyes was enhanced by introduction of thiophene units, as vinylene units proved problematic for dye aggregation, achieving efficiency of up to 8.1% [17, 143]. Furthermore, addition of more cyanate groups to the π bridge of the dye resulted in additional extension of the absorption spectrum [144]. Indoline dyes showed a broader absorption spectrum after incorporation of a rhodamine framework, reaching efficiency of 9.5% after optimization of the dye structure and accompanying ETL [145–147]. One of the highest near-infrared IPCE values for organic dyes of donor-π-acceptor structure was achieved using a tetrahydroquinoline dye, through modification to separate the anchoring and acceptor group, resulting in device efficiency of 3.7% [148]. The broadest IPCE spectrum of organic dye-based DSSCs reached 920 nm and was achieved by incorporation of a phenoxazine dye with a thienyl π bridge and a corhodanine acceptor [149]. Other organic dyes that have been engineered for red-shifted absorption include N,N-dialkylaniline dyes which reached 6.8% efficiency [150, 151], squaraine dyes which show good absorption in the near-infrared region [17, 152], and boradiazaindacene dye, which has a near-infrared absorption spectrum (600–800 nm) and produced a device of 1.7% efficiency [153].

Despite the extensive work that has been done on this aspect of DSSCs, an optimum sensitizer that can cover the entire solar spectrum has not been fabricated yet, although mixing of different sensitizers has been implemented in devices with impressive results [18, 57]. It has been estimated that by extending the absorption spectrum to 920 nm, PCE of 19% can be achieved [154].

2.1.4.2. Leakage in DSSCs

2.1.4.3. Photo-bleaching in DSSCs

Implementation of QDs offers a variety of benefits over dyes as sensitizers in a solar cell. Perhaps their biggest benefit is the ability to tune their spectral properties easily by varying the diameter of the QD, due to quantization effects present. By modifying the energy levels of the QDs, light absorption and electron injection can be tailored to match the needs of the solar cell [171, 172], whereas generation of multiple excitons by a single photon and hot electrons offer new possibilities for enhancement of device performance [171, 173].

Good QDSSC operation requires that the conduction band of the QD sensitizer be higher than the conduction band of the ETL for efficient electron transfer between the two, as the offset provides the driving force for charge transfer [163, 171]. While the open circuit voltage of the device is independent of the size of the QDs used in the sensitizer, as demonstrated in figure 5(a), the device photocurrent is directly correlated to the properties of the QDs [171]. It has been observed that smaller QDs result in higher photocurrent due to the higher conduction band as shown in figure 5(b), implying that a larger driving force will be available for electron injection to the ETL [171, 174]. However, decreasing QD size is also associated with more limited response in the visible part of the spectrum as shown in figure 5(c) [171].

Zoom In Zoom Out Reset image size

The emission decay of CdSe/TiO₂ electrodes was used to estimate electron transfer constants through measurement of the recombination rate constants [170, 171]. Exponential decay models were implemented to extract emission lifetimes, and it was observed that for smaller nanoparticles, shorter electron transfer constants are obtained [171]. Furthermore, there was no apparent difference between the QDs on TiO₂ nanoparticles and TiO₂ nanotubes, indicating that the morphology of the ETL does not influence the charge injection process. This further supports the evidence that photocurrent is dominated by QD properties. IPCE was also seen to improve with smaller QDs, establishing that since they possess a more energetic excited state, their electron injection rate is higher. The morphology of the ETL becomes important where open-circuit voltage is concerned. As seen in figure 5(a), the photo-voltage of the solar cell decays more slowly after illumination stops for the TiO₂ nanotube substrate, which is direct indication that the electrons injected in the TiO₂ have a longer lifetime before recombination, hence the tubular structure can be used to minimize losses at grain boundaries [163, 171]. Therefore, both smaller and larger QDs offer different benefits for the solar cell and its operation. Thus, a tandem structure ('rainbow' configuration) has been proposed for QDSSCs, which would leverage benefits from both small and large QDs [163, 171]. A gradient assembly of QDs of various sizes could be used, in which small QDs would absorb light of higher energy, allowing longer wavelengths to be transmitted and eventually be absorbed by larger QDs below them. This would utilize the fast electron injection of small QDs and high absorption of large QDs at the same time [171].

This concept has also encouraged research into alloyed QDs which would combine the beneficial properties of various materials to fabricate QDs that are most suitable for QDSSCs [163, 175]. This has led to the preparation of the champion device with 11.61% certified efficiency, which used 4 nm Zn-Cu-In-Se QDs as sensitizers and combined absorption onset of 1000 nm and fast electron injection rate of 9.1 × 10¹⁰ s⁻¹ [170].

Another major benefit of QD sensitizers is the potentially higher stability of QDs in oxygen and water, in comparison to molecular dyes [176]. While originally the iodide/triiodide couple was implemented in QDSSCs as well, it was observed to be detrimental to QDs which were unstable in this electrolyte [163, 165]. Instead, the sulfide/polysulfide system was implemented as redox couple, with good results [177]. Nevertheless, this has shown issues with low open-circuit voltage due to the high redox potential of the polysulfide electrode, and various techniques to address this issue have been used, including adjusting the concentration of the redox couple, changing the solvent composition, and incorporating additives [178–181]. However, the inability of the polysulfide electrolyte to scavenge holes from QDs efficiently has been a source of device instability, as the QDs oxidize and subsequently degrade [175]. This has encouraged research into alternative electrode materials, which resulted in improved device fill factor and enhanced stability [165, 175, 182]. Such materials include Cu₂S or Au, instead of the typical Pt. Furthermore, QDs with enhanced stability have been engineered to address this issue. Alloyed QDs are observed to demonstrate increased stability in the polysulfide environment in comparison to QDs prepared with the constituent materials [175, 183].

Like DSSCs, QDSSCs also face instability issues due to leakage of the liquid electrolyte [170]. Research efforts to implement quasi-solid and solid-state electrolytes in QDSSCs, such as ionic liquids, gels, hydrogels, spiro-OMeTAD, and P3HT have met with great success [184–188]. QDSSC with a gel electrolyte combined with CdSe_xTe_1−x QDs has demonstrated power efficiency of 9.21% [189], whereas a device with CuS QDs and P3HT has produced PCE of 8.07% [190]. QD-sensitized ETLs appear to form better junctions with solid-state electrolytes as compared to dye-sensitized ETLs, as the thinner layer of sensitizer alleviates the issue of insufficient penetration of the solid-state electrolyte into the sensitizer/ETL structure that is commonly seen in DSSCs [164, 191]. Consequently, more stable solid-state devices are feasible when QDs are used as the light harvesting medium, instead of dyes.

Another bottleneck QDSSCs must overcome before the PCE can be enhanced is the poor loading of QDs on the mesoporous or nanostructured ETL. This has been associated with enhanced recombination of photogenerated excitons in QDs at the interface, thereby limiting device efficiency [191, 192]. Different techniques to prepare and deposit QDs on the ETL have been developed, including techniques that QDs are grown in situ and ex situ [193–195]. The former encompasses techniques such as chemical bath deposition, which is a fast and easy process [163, 193], as well as successive ionic layer adsorption and reaction (SILAR), which allows easy tuning of QD sizes [163, 194]. QDs grown ex situ are deposited onto the ETL with the aid of molecular linkers to enable attachment of the QDs to the metal oxide ETL [163, 195]. Other methods include electrodeposition [196], spray pyrolysis deposition [197], and pulsed layer deposition [198].

Work has been done to optimize these techniques and find experimental conditions for optimum loading of QDs on the ETL. In the case of QDs grown ex situ, mercaptopropionic acid (MPA) and cysteine have been identified as good molecular linkers for attachment to metal oxides [199, 200]. It was observed that as coverage of the ETL increased, the IPCE was initially enhanced, before reaching some threshold value after which it started decreasing, which was attributed to enhanced recombination of photogenerated charge carriers brought on by aggregation of QDs as the electrons must now transfer across QD/QD interfaces before reaching the metal oxide [200]. However, the amount of QDs that cause the decline in IPCE is dependent on the molecular linker used, indicating that the choice of linker is influencing the electron transfer process [200, 201]. Similar trends have also been observed in samples with QDs deposited by SILAR or CBD [202–205]. These deposition methods also result in the growth of larger QDs as the quantity of QDs increases. Atomic layer deposition (ALD) has been suggested as an alternative method for in situ growth of QDs that does not suffer from the same problem [192]. This is another aspect in which smaller QDs appear to be superior. The QDSSC with record PCE was obtained with 4 nm QDs, and the authors have attributed the high efficiency partially to the improved coverage of the TiO₂ ETL due to the small QD size [170]. The small diameter QDs reduce the blocking effect observed otherwise, allowing the QDs to penetrate through the ETL [175, 206, 207].

Another common issue observed in QDSSCs is poor quantum yield of electron and hole injection into the respective adjacent layers. This has been associated with trap state defects in QDs which result in charge recombination [170, 206, 208, 209]. It has been found that deposition of a 2–3 nm thin ZnS layer on the photoanode improved device performance as it passivated surface traps and prevented electron injection to the HTM [165, 204, 210], whereas alloyed QDs could be engineered to minimize the amount of surface traps [170, 211, 212]. The choice of passivating ligands on QD-sensitized metal oxides has enhanced internal quantum efficiency of photon-to-electron conversion to ∼100%, matching that of DSSCs [213]. Furthermore, use of small QDs increases the difference between excited state of QD and ETL conduction band, providing a larger driving force for electron injection, and leading to higher IPCE [171]. Deposition on nanostructured TiO₂ (nanotubes) was also observed to enhance electron injection as it reduced recombination losses observed at grain boundaries of TiO₂ nanoparticles [171]. QDSSCs additionally offer the potential of leveraging impact ionization occurring in QDs, which would result in two or more excitons forming with a single photon, as well as hot electron transfer [163, 214, 215]. Work done on this front has resulted in impressive outcomes, with absorbed photon-to-current efficiency of PbS QD-sensitized TiO₂ exceeding 100% [214].

Pb-based QDs, in particular PbS and PbSe, have been very popular choices as sensitizers in colloidal QD solar cells. PbS QDs were used both in the first case of colloidal QD solar cells, as well as high efficiency colloidal QD solar cells reaching 8.55%, because of their excellent absorption properties, which can be tuned from 800 to 2000 nm [222–224]. Furthermore, PbS QDs demonstrate good air stability, which has been a challenge for colloidal QD solar cells. PbSe QDs have also been implemented in solar cells with impressive results, and both PbS and PbSe have demonstrated multiple exciton generation resulting in internal quantum efficiency that exceeded 100%, proving that QD-based solar cells could reach high efficiencies, beyond that of DSSCs [214, 225]. Hot electron transfer from PbSe QD films to TiO₂ has also been observed, which can be additionally leveraged for high efficiency devices [226].

Since one of the major limitations of colloidal QD solar cells is charge transport within the QD layer, the conduction characteristics of materials such as PbSe, PbS, and CdSe QD films were investigated to elucidate the processes occurring during device operation. It was found that after exciton dissociation, charge carrier transport occurs through phonon-assisted hopping between the energy levels in QDs, contrary to earlier claims of band-like conductivity [227, 228]. PbS and PbSe QD films have been shown to possess good electrical properties, with PbSe QD films reaching mobilities of 7 cm² V⁻¹ s⁻¹ after treatment with passivation and infilling [229], which will be re-visited in the next sections.

The charge carriers inside the QD film are driven by drift currents generated by the electric field in the device [217]. Since carrier transport is below optimal, improvements are needed to enhance the drift of carriers. Efforts have been focused on band engineering and ligand modification to improve charge transport.

Ligand modification has proven key in controlling the electrical properties of QD films. Use of ethanedithiol (EDT) molecules to assist in film fabrication has demonstrated improvement in carrier mobility, associated with reduced inter-particle spacing [230, 231]. But owing to the vulnerability of organic ligands, inorganic ligands were developed to enhance electronic transport and passivate surface defects [232]. Metal chalcogenide complexes were found to result in electron mobilities of 16 cm² V⁻¹ s⁻¹ in CdSe QDs [233], whereas atomic ligand passivation produced a PbS QD device with efficiency of 6% in the first attempt to fabricate a photovoltaic device with inorganic ligands [232]. In this device, the PbS QDs were capped with oleic acid ligands when synthesized, with thiol treatments removing the oleic acid and passivating the QDs through a Pb–-S bond [232, 234]. The PbS QDs were then treated in CdCl₂, tetradecylphosphonic acid (TDPA), and oleylamine (OLA), which improved sized distribution, stability, and reduced defects. Finally, the film was treated with either cetyltrimethylammonium bromide (CTAB), hexadecyltrimethylammonium chloride (HTAC), or tetrabutylammonium iodide (TBAI) to generate halide-passivated PbS QD films.

Apart from ligand modification, band engineering has also been very useful in improving electrical transport in colloidal QD solar cells, involving tuning of the band alignment between the colloidal QD film and the ETL to favor electron injection into the ETL [217]. This has also led to the concept of quantum funnels, essentially the preparation of graded band gap using five QD layers and leveraging the fact that carriers will end up in the layer with the smallest band gap. This has resulted in improved fill factor, open-circuit voltage, and short-circuit current [217, 235, 236]. In addition, graded doping architectures have been used to improve device efficiency [237]. The dopant can be used to tune the conduction band edge and band bending at interfaces, with ligands such as 3-mercaptobutyric acid and EDT used to dope PbS QD films, showing impressive results and leading to record efficiencies [224, 235]. Other efforts have included doping of the metal oxide used as ETL, incorporating impurities in TiO₂ and magnesium doping of ZnO, to achieve optimal band offset [238, 239].

Another approach to address the limited mobility and conductivity demonstrated by colloidal QD films is infilling with metal oxides. Deposition of Al₂O₃ and ZnO via ALD to fill the pores of PbSe QD films was observed to reduce the inter-QD tunnel barriers, enhancing carrier mobility [240, 241]. Enhanced conductivity and carrier mobility was also observed after infilling of CdSe QD films with ZnO [242]. ALD appears to be the optimum technique for deposition of the infilling material as it can uniformly coat structures characterized by high aspect ratio, and in work done on infilling a PbSe QD film with amorphous Al₂O₃, high charge mobility was achieved (7 cm² V⁻¹ s⁻¹), due to passivation of the QD surface trap states, as well as strong electronic coupling of the sodium sulfide ligands [229]. Furthermore, infilling of a PbSe QD film with Al₂O₃ or Al₂O₃/ZnO was observed to result in multiple exciton generation in the QDs, which was not observed in the non-filled films, providing additional motivation for infilling [243]. This is a consequence of the enhanced charge mobility in the infilled films which reduces the possibility of Auger recombination of charges.

Since the presence of defects hinders charge carrier transport through the QD film and enhances the possibility of charge recombination before they reach their respective contacts, surface passivation to eliminate these trap states can be a useful tool in improving device performance [227]. This can be achieved through different techniques, with ligand passivation being a common method. It was observed that PbS QD films that were treated with 3-mercaptopropionic acid (MPA) demonstrated improved device performance as compared to oleic acid or EDT, associated with superior passivation of defects, leading to higher mobility and minority carrier transport [244]. Infilling is also used to yield improved device stability which is achieved through surface passivation, as it prevents oxidation and photothermal degradation of QDs exposed to ambient conditions [229, 240, 243].

Surface passivation is achieved by exposure of PbSe QDs to chlorine gas, as it results in a thin PbCl₂ layer which passivates the QDs and prevents oxidation [245]. PbCl₂ can also be used as a precursor for PbSe QDs to create a passivating thin PbCl₂ layer [246, 247], or alternatively, treatment of PbSe QDs with NH₄Cl can produce a thin PbCl₂ layer [248]. In addition, cation exchange techniques have been used to introduce metal cations post-synthesis and enable implementation of techniques used for Cd structures to be used in Pb structures that can ultimately be used to improve carrier transport and reduction of trap states [249, 250] which has been used to passivate PbSe QD films and eventually produce air-stable devices [250].

While many improvements are needed in terms of device stability, conductivity, and performance, colloidal QD solar cells have demonstrated impressive improvements in efficiency since their inception in 2005, reaching efficiencies close to those of DSSCs [221], showing that they are a very promising photovoltaic technology.

Scattering of light in NWs is described in terms of wave optics and depends on NW diameter, length, pitch and slant angle [270, 273, 274]. By warrying these parameters it is possible to improve scattering and absorption within a material compared to its bulk counterpart. It is also important to remember that the dependence of these parameters vary with different materials due to differences in their electronic band structure and dielectric properties. Li et al used simulations based on full wave finite element method to show that the hexagonally arranged Si nanopillars with pitch and diameter lying close to 600 nm and 500 nm respectively has increased absorption compared to thin films with the same amount of material [275].

Apart from light scattering, the presence of guided leaky resonant modes are also responsible for increased absorption in nanowires [274, 276]. Multiple studies using analytical waveguide theory have shown that the resonance position of these modes are minimally perturbed from a single NW to sparsely spaced III–V NW array [274, 277], while near field coupling between neighboring NWs is significant in Si NWs [278, 279]. Calculation from Anttu et al showed that these modes had strong diameter dependence and weak dependence on pitch and length in InP NWs with minimal absorption below 100 nm diameter and maximum at 170 and 410 nm respectively [274]. The increase in absorption at these diameters is due to the strong excitation of HE₁₁ (170 nm) and HE₂₂ (410 nm) modes by incident light followed by strong absorption by the NWs. The design of the simulation along with results of which are shown in figure 8. In addition, Dhindsa et al reported that smaller NWs have low absorption at longer wavelengths irrespective of packing density [280], further confirming the dependence of nanowire diameter on performance.

Zoom In Zoom Out Reset image size

Another important property of NWs is the ability to separate charge carriers near the p–n junction. NWs have significant advantages over bulk materials as the charge carries have to transverse much smaller distances compared to their diffusion lengths, minimizing losses due to scattering and radiative recombination [281, 282]. Some of the important parameters that determine electrical properties of NW solar cells are doping concentration, geometry and surface passivation. In III–V NWs, crystal structure also seem to play an important role in charge separation [264]. Since III–V's can exist in zinc blend and wurzite crystal structure or predominantly as a mixture of the two, such two phase systems can act as scatterers and trap states for radiative recombination [283, 284].

Incorporation of dopants during the growth step on nanowires to create p and n type semiconductor is another important parameter determining the electrical properties of nanowires [285]. In vapor–liquid–solid (VLS), which is the most common growth method for Si NWs, the dopants are introduced as vapor precursors [285]. While the effective incorporation of dopants during growth is still being investigated, the present understanding is that the dopants enter via the seed particle-nanowire interface [285]. In 2009, Nduwimana et al used first principle density functional theory to show charge carrier densities in p and n type doped Si nanowires [286]. Further, Murthy et al used time resolved measurements to show efficient charge carrier separation in Si NWs by having a radial doping gradient [287]. Similar ultrafast charge carrier dynamics studies were also done on GaAs [288] and InP [288, 289] NWs to understand the effect of doping. Some of the recent research on the properties of NWs include charge separation in VLS grown axial nanowires [290], effect of Nb doping and oxygen vacancy centers in TiO₂ [291] and ZnO [291–294] respectively.

Due to limitations in fabrication, even with extensive theoretical and experimental study on nanowire optical enhancements and electric properties, it was tough to achieve similar results with device performance [295]. But towards the second half of the last decade, multiple experimental groups have made progress in fabricating structures with results in agreement with theoretical predictions. An image of a working device along with its J–V characteristics are shown in figure 9.

Zoom In Zoom Out Reset image size

Some of the nanowire fabricating techniques will be discussed in brief in the following section, prior to which we need to look the two designs in which we can fabricate a nanowire p–n junction. p–n junction nanowires can be fabricated in both radial and axial direction as shown in figure 10.

Zoom In Zoom Out Reset image size

In 2012, it was shown experimentally that radial designs are much less prone to surface defect compared to axial designs though both are capable of maintaining a V_oc greater than 0.7 V [296]. It was further confirmed that axial designing of GaAs exhibits large leakage current under backward bias [297]. A radial p–n junction is also ideal for orthogonalization of light absorption and charge carrier collection [281]. However, fabricating a radial NW heterostructure requires precise lattice matching steps and brings up the production costs, while axial designs are passive to any such lattice mismatch strains. Some of the earlier work on coaxial nanowires in 2007 achieved a quantum efficiency of 12% at 690 nm incident light [265] and a PCE of 3.4% in a coaxial p–i–n coaxial silicon solar cell [258]. However, one of the first radial p–n junction solar cells reported a PCE of 12.2% [298] followed by a current density of 40 mA cm⁻² in radial Si p–i–n nanowire solar cell [299]. More recently, a hybrid solar cell of Si nanowires and PEDOT:PSS showed a PCE of 13.2% [300]. Apart from Si, there have been various experimental realizations of III–V nanowire solar cells. In 2013, radial GaAs nanopillar solar cells with 70% quantum efficiencies and 63% fill factors were reported [301]. Even though axial design has a disadvantage over radial design, some of the highest recorded efficiencies have been reported for axial nanowires. Axial GaAs nanowire solar cells in 2015 demonstrated a record efficiency of 15.3% [269] in well over the previous highest value of 13.8% in InP nanowire solar cells [302]. This value was exceeded in coaxial InP NWs with the help of self-aligned nanoparticles which helped in improved absorption, with a reported a PCE value of 17.8% [251]. Similar high performance solar cells have also been reported for CdS [303–305], ZnO [306, 307], Cu₂O [306] and more recently, for perovskite nanowires [271]. Research has also begun into tapered NW structures like nanocones [308] and dual-diameter tandem NW solar cells [309]. Such structures help reduce reflectance while maximizing effective absorption.

For the realization of efficient light management and enhanced PCE in nanowire solar cells, it is of paramount importance to develop low cost, precise and reproducible fabrication routes. In the section we will look at different growth methods employed for nanowires (mostly Si and III–V's). Since several detailed reviews on nanowire growths are available [310–313], the discussion is intended as a brief overview with some recent advances made.

VLS method is a growth mechanism in 1D materials known for than more than 50 years [314] and is the most common technique used for the growth of Si and other semiconductor nanowires. They are cheap and can grow single crystal nanowires with high crystallinity [281, 282]. The process involves passing a vapor of the precursors on a substrate with a solid seed nanoparticle, generally gold, at very high temperatures (500 °C–1000 °C) [315, 316] in the presence of a liquid catalyst. The liquid catalyst is formed during the initial phase, when the precursors dissociate to form a metal-nanoparticle droplet, that on further addition of dissociated precursor, leads to nucleation sites at its surface. SiH₄, disilane, SiH₂ and SiCl₄ are the most commonly used source gasses for Si [317]. In 2009, growth of GaAs nanowires was demonstrated via VLS in a gas source molecular beam epitaxial system [295]. Another advantage of VLS is the ease of growing p–n heterostructures by passing different doping gases along with the source gases. One of the earlier uses of VLS growth mechanism was to grow Si nanowires using chemical vapor deposition (CVD) technique [316]. They further demonstrated patterned growth of nanowires using Au colloidal seed particles. CVD [315, 316] and molecular beam epitaxy techniques [318, 319] are two of the common fabrication routes that uses VLS mechanism for aligned single crystalline growth of nanowires. The 17.8% efficiency reported in nanowires were grown using VLS [251]. However, it must be noted that multiple other factors like plasmonic enhancement played an important role in the high efficiency reported.

Metal organic chemical vapor epitaxy (MOVPE ) is predominantly used for III–V nanowire growth as initial results of GaAs nanowire solar cell grown using VLS showed very low efficiencies, less than 1% [295]. Though multiple reports on InP nanowire growth using MOVPE existed [320–323], the first InP nanowire solar cell, based on the method was reported in 2009, and showed an impressive PCE of 3.37% [324]. In the case of GaAs, it was not until 2012, that an impressive PCE of 2.54% was reported in densely packed uniform GaAs nanowires [301]. Further, the same group improved on the efficiency to 6.63% in 2013 [325] by making use of surface passivation techniques. Since then, multiple reports showing high efficiencies in GaAs [252, 268] InP [251, 302] have been reported. The MOCVE technique involves a combination of electron beam lithography and vapor epitaxy to grow nanowires of uniform thickness and shape [301]. The first step involves deposition of lithographically patterned SiO₂ mask with an array of holes onto a p-type substrate of the nanowire material. This is followed by chemical vapor epitaxial growth of the nanowire though the etched out holes, followed by growing the n-type shell on the p-type nanowires, by varying the growth conditions. The process involves elevated temperatures in the range of 600 °C–800 °C. Figure 11 shows InAs nanowires gown using MOVPE.

Zoom In Zoom Out Reset image size

Most of the growth techniques mentioned above are based on bottom-up methods. Many top down synthesis routes also exist for the fabrication of nanowires, although, challenges exist due to the need for high precision and throughput production. In 2005 a growth technique was developed for single Si nanowire using aqueous electroless etching method, involving etching a Si wafer using HF acid followed by deposition of a-Si using disilane gas at 450 °C [326, 327]. This synthesis technique was made use of to test a Si nanowire solar cell which reported a low efficiency of 0.5% [328]. The most common type of wet etching—metal assisted chemical (MAC) etching involves a noble metal nanoparticle covering the Si wafer followed by reaction with an etchant [329–331]. The etchant facilitates oxidation of the Si wafer beneath the metal and as the process progresses, the Si beneath the metal nanoparticle is etched away forming Si nanowires [282]. This method is cheap, and can be used for mass production of isotropic nanowires. The etching technique was further improved upon with the addition of template assisted MAC etching [327, 332] as shown in figure 12. This technique offers better control over density, shape and length of the nanowires. MAC technique is still popular for the growth of nanowires and results have shown comparable solar cell efficiencies [333, 334]. Apart from wet etching, a number of dry etching techniques such as laser ablation [335], reactive iron etching [336], plasma etching [337] and chemical assisted ion beam etching [338] also exist for nanowire growth.

Zoom In Zoom Out Reset image size

The above discussed methods are the common techniques for growth of nanowires, but there have also been reports of various improvements on existing techniques to show high efficiencies. MOCVD was used to report a PCE of 8.8% in graphene/single nanowire Schottky junction solar cell [339, 340]. Nanowire solar cells with efficient light management has the capability to outperform their bulk counterparts both in efficiency and cost. Due to fabrications limitations, material quality and poor design they are yet to achieve efficiencies close to their theoretical predictions. But the last few years have seen substantial improvement in nanowire solar cells and shows great promise to improve further in the next few years.

3.1.1.1. Far field scattering

3.1.1.2. Localized field scattering

3.1.2.1. Hot electron injection (HEI)

In addition to enhancement through efficient absorption, recent investigations proved that it is possible for plasmonic nanostructures to transfer hot electrons from the metal directly into the conduction band of the semiconductor [164]. Hot electrons refer to electrons that have gained kinetic energy from an external electric field and are no longer in thermal equilibrium with the atom, hence are described by Fermi statistics with an elevated temperature. Schematic and band diagram of hot electron transfer is shown in figure 15 [343]. Many existing PV technologies such as DSSC, quantum dot solar cells and quantum dot-sensitized solar cells have previously utilized HEI to further improve their power conversion efficiencies [164, 408, 409]. Around 2004–2005, several groups reported on the ability of plasmonic nanostructures to generate 'hot electrons' under illumination, and these led to major progress in photo-catalysis and photovoltaics [346–348]. Hot electrons in metal nanoparticles are generated as a result of non-radiative decay of localized surface plasmons. LSPR excitation decay occurs on a time scale of femtoseconds with the emission of radiation or non-radiatively by transferring the energy to hot electrons, which can undergo photoemission, if its energy exceeds the work function of the material. A fraction of the energy is also lost as heat, but as mentioned above, ohmic losses are lower in Au and Ag. Probabilistically, majority of excited electrons originate from intraband excitations within the conduction band compared to interband excitations, as the d band energy levels lie 2.4 and 4 eV below Fermi level in Au and Ag, respectively [343].

Zoom In Zoom Out Reset image size

The excited hot electrons can thus escape the nanoparticle and can be injected into the conduction band of the semiconductor by forming a Schottky junction. Note that this mechanism becomes efficient only if the energy needed for such injection is lower than the band gap of the active material. Even with a perfectly absorbing plasmonic nanostructure, the PCE enhancement through HEI is limited to <8% [411]. However, modifying the electronic density of states of the donor and acceptor could help improve upon this value.

One of the first comprehensive studies of HEI came from observation of photo-induced electron transfer between Au and thin film and nonporous TiO₂ [346]. Shortly afterward, multiple results observing electron diffusion between metal nanoparticles and TiO₂ were published [412–415]. Some of the recent results of HEI in PV include using HEI in sandwiched TiO_x–Au–TiO_xto improve efficiency in perovskite solar cells [416], improving NIR absorption in Si solar cells using Ag nano arrays [417], independent plasmonic solar cell with 0.2% PCE [418] and solid state ITO∣Au-NPs∣TiO₂ solar cell with 5.84% photon to electron conversion rate at 700 nm incidence [419]. HEI using plasmonics have shown the ability to increase population of electron hole pairs in solar cells, but their performance have been limited by the fraction of the hot carriers having energy to cross the Schottky barrier and various hot electron relaxation pathways [420].

3.1.2.2. Plasmon resonance energy transfer

UC is an anti-Stokes process, can be done at low power, does not require highly coherent light and can be achieved even using Xenon or Halogen lamps [434–436]. The underlying principle of up conversion can be understood from figure 18. Energy is absorbed from the ground state and electrons are carried to their first exited state. The first exited state is a metastable state and require a relatively long lifetime for the UC process to work. Consequently, another photon is absorbed and now the highly populated metastable state is exited to the higher energy state, followed by radiative recombination to the lowest energy state [429, 437]. During the whole process two photons of lower energy were absorbed and a photon of higher energy was emitted.

Zoom In Zoom Out Reset image size

The most common materials used for up conversion processes are lanthanide nanoparticle based luminescence [427–430] and triplet triple annihilation based on organic molecules [427, 439]. In this article, we will focus on the lanthanide nanomaterials for up conversion. Although the basic energy scheme remains the same, there are multiple ways to achieve up conversions and the most effective way is to use an energy transfer up converison, in which a host material is responsible for transferring energy to the activator nanophore. The first energy transfer excites the activator to the metastable state, this can either be followed by another energy transfer or the activator absorbs the next photon to reach the excited state [440, 441]. Similarly, there are other pathways to achieve such energy transfers [442, 443]. The advantage of using energy transfer up conversion is that we do not require a high concentration of activator ions [440].

The host is generally an inorganic molecule with minimum non-radiative losses. Researchers have studied various host matrices including fluorides [444–446], other halides [447], oxides [448, 449], and phosphates [450, 451] etc. For lanthanide nanoparticles, NaYbF₄ is the most common used host. However, using Gd₂O₂S [452–454], BaY₂F₈ [455, 456] and fluorozirconate glass [457, 458] as host material is also common. Er³⁺ [431, 434, 445, 452, 454–456, 459], Tm³⁺ [447, 450, 460], and Ho³⁺ [460, 461] ions, with their appropriate f electron configurations, nearly evenly spaced energy levels and long excitation lifetime, are commonly used as the activator material. The three step UC in Er³⁺ energy levels are shown in figure 18 [438]. The larger absorption cross sectional area in Yb³⁺ ions compared to activator along with their convenient relaxation energy, which resonates with the excitation energies of activator ions are major factors in improving the efficiency of the energy transfer UC [436].

The first proposed use of UC was in 1982 using terbium-doped lanthanum fluoride and thulium-doped calcium tungstate [437], followed by the first demonstration in 1996 of an improvement in GaAs solar cell performance using co-doped Er³⁺ and Yb³⁺ [462]. Over the years, different solar calls have been tested using various lanthanides as dopants. In 2002, a theoretical prediction was made of a PCE of 47.6% in c-Si solar cells using sub-band gap light, considerably higher than the Shockley–Quessier limit [463]. Erbium-doped sodium yttrium fluoride (NaYF₄:Er³⁺) as the host-activator matrix achieved an internal quantum efficiency of 3.8% in c-Si solar cells at laser excitation of 1523 nm [459]. In 2009, Gd₂(Mo0₄)₃:Er³⁺ was proposed as an up conversion phosphor for Si solar cells [464], but it was only in 2010, that UC was finally tested in organic and DSSCs, with the use of Er³⁺,Yb³⁺ co‐doped LaF₃‐TiO₂ nanocomposites as a middle layer in DSSC to produce an improved PCE of 2.69% [465]. Yttrium fluoride host doped with erbium in P₃HT:PCBM solar cell showed a four-fold increase in photocurrent at 975 nm illumination [466]. With silicon solar cells, some of the recent works involve improved external quantum efficiencies at long wavelength irradiance and approaches into NIR broadband UCs [467, 468].

Some of the recent works on up conversion in PV are summarized in table 1.

Losses due to thermalization can be reduced with the use of DC materials that help in generation of multiple electron–hole pairs for incident photons of energy greater than twice the band gap of the material. Hence, DC is more beneficial for solar cells with smaller band gaps as it converts a photon of higher energy to two photons of lower energy, implying an external quantum efficiency of 200%. Although a DC layer is typically applied to the front surface of a solar cell, it is possible to add a rear layer in the case of bifacial solar cells [427, 428]. It was theoretically predicted that a conversion efficiency of 38.6% can be achieved in c-Si solar cells attached to an ideal DC convertor under un-concentrated light, by increasing the available spectral irradiance by 32% [475].

DC works in materials with an intermediate state between their bands. It was shown that in such materials, under low light intensity such as concentrate light, absorption occurs via band to band transition, while radiative recombination occurs with the help of an intermediate level [475]. However, another way to achieve such DC is through cooperative energy transfer between ions, achieved through co-doping two different ionic systems in a material [476, 477].

In 2005, a cooperative energy transfer efficiency of 88% was reported in (Yb_xY_1−x)PO₄ doped with 1% Tb³⁺ upon excitation using a Xe lamp [478]. As expected, the energy transfer in DC occurs opposite to that of UC, that is, from two Tb³⁺ to one Yb³⁺ ion. Around the same time, multiple groups also reported improved PCE and short current density in c-Si solar cells with the use DC materials [478]. Broadband DC of UV light resulting in emission of NIR photons upon excitation using was reported in 2008 with a high energy transfer efficiency of 74% between Ce³⁺ and Yb³⁺ ions co-doped in borated glasses [479]. Further, they reported a DC of visible light at 475 nm in a co-doped system of Yb ³⁺ doped oxyfluoride glass ceramics with precipitated TbF³ nanocrystals [480]. Other commonly used materials for down conversion are CdO nanotips [481], ZnO:Er³⁺/Yb³⁺ [482], GdAl₃(BO₃)₄(Pr,Tb)₃₊,Yb³⁺ [483], YF₃ doped with Pr³⁺ Yb³⁺ [484], Gd₂O₂S:Tm³⁺ [485], and lanthanide ion doped glasses [486]. Many groups have reported improved performance in solar cells with DC materials, and some of the recent results are shown in table 2.

Over the last decade active research on UC and DC materials have helped in improving the efficiencies of solar cells. However, the reported values are still low and demands testing new materials and better fabrication techniques. Researchers have also tested plasmonic and photonic materials as efficient UC/DC materials [487, 488]. These materials have additional benefits of enhanced scattering and other plasmonic effects described earlier. Another approach is the use of concentrated light source [427, 489]. UC/DC are yet to achieve substantial improvements that can be implemented into existing commercial devices. However, over the last decade, encouraging results with an increasing trend in efficiencies is visible.