Light harvester band gap engineering in excitonic solar cells: A case study on semiconducting quantum dots sensitized rainbow solar cells

Introduction

Semiconducting sensitized solar cells are promising devices for solar energy conversion that have gained a remarkable attention over the past years [1]. They belong to the larger class of excitonic solar cells (XSCs), which are photoelectrochemical cells exploiting as (photo) anode a wide band gap metal oxide-based photoanode, typically made by TiO₂ nanoparticles. The anode is then sensitized with a light harvester, whose role relies in generating an electron-hole pair under irradiation, borrowing the concepts exploited by green plants in producing energy under solar irradiation [2].

XSCs have gathered considerable interest as alternative devices to silicon technology, due to several benefits, such as the use of relatively cheap materials, the applications of low cost processes, the capability of taking advantage of diffuse light and the possibility to apply light harvesters with different absorbing features, thus enabling the exploitation of broader region of the solar spectrum, possibly including the near infrared (NIR).

Quantum dots sensitized solar cells (QDSSCs) exploit the architecture of the most famous dye sensitized solar cells [3], while using semiconducting quantum dots (QDs) as light harvesters. The particular attention paid to these systems is motivated by the outstanding opto-electronic properties featured by QDs, especially related to their tunable band gap, possibility of direct injection of hot electrons before thermalization, remarkably high molar extinction coefficients and large dipole momentum [4]. Thanks to the claimed multi exciton generation effect and the presence of intraband transitions [4, 6, 7], expectations on the possibility to overcome the Queisser-Shockley limit of photoconversion efficiency (PCE) have even arisen.

Despite these promises, QDSSCs still feature performances lower than their dye counterparts. This is due to several unfixed issues, among which it is worth reminding a certain difficulty in colloidal QD uptaking on metal oxide scaffold, the lack of reproducible procedures for the fabrication of effective counter electrode [8], a relatively low stability of light harvesters due to photocorrosion [9]. Colloidal QDs retain a great appeal due to the tunability of optoelectronic properties, chemical environment surrounding the inorganic nanoparticles and composition of nanocrystals, allowing in principle for very good management of anchoring to metal oxide scaffolds constituting the photoanodes together with modulation of solar spectral region of absorbance. Recent efforts in this frame had resulted in improved device functional perfomances. Particularly worth mentioning is the paper by Zhong and coworkers [10], in which Authors present the synthesis and application as light harvester of alloyed CdSe_x Te₁–x QDs with excitonic onset at 800 nm (providing for a photoconversion efficiency higher than 6 %).

Recently [11], the use of naked QDs, directly generated on the metal oxide scaffold, has proved as effective strategy to fabricate relatively high performance devices, with the additional advantage to easily produce hybrid network composed by different materials, allowing the fabrication of so called rainbow solar cells. This latter point is of particular relevance in view of effectively exploit the near infrared region of the solar spectrum, which is currently not only lost in traditional photovoltaics, but even detrimental, reducing the performances by heating. The fabrication of engineered QD networks through SILAR (successive ionic layer absorption and reaction) has gained fame in the field of QDSSCs since it enables a very simple way to uptake polydispersed QDs on metal oxide (MOX) scaffolds by directly generating and growing the nanocrystals on a photoanode surface. These QDs are naked, i.e., they are provided with no molecular ligand, as usually applied for colloidal nanocrystals, which pose problems for both loading strategies (much longer times are needed – 24–72 h by uptaking them through the ligand exchange reaction – not to mention the time required for the ex situ synthesis) and electron transfer rates [12–14]. When SILAR approach is instead applied, photoanode sensitization lasts less than 1 h and an intimate contact between the light harvesters and the electrode is obtained. This configuration should favor exciton separation and charge injection from the QD to the oxide photoanode. Amount and sizes of QDs can moreover be easily controlled by modulating the number of SILAR cycles, which results in a very simple strategy for band gap tuning [11, 14].

Most commonly exploited QDs by SILAR are Cd and Pb sulfide and selenide, which are able to harvest light from about 400 nm up to about 900 nm; very recently, a few attempts to fabricate more complex materials via SILAR, among which doped QDs [16], mixed CdSe_x S₍₁–x₎ [24], PbS:Hg QDs [17] have been also tried with rather good success. The remarkable efforts lavished over the last few years (the first paper on SILAR-sensitized solar cells appeared in 2009) have succeeded in improving the photoconversion efficiency up to a 6 % and expanding the light collection toward the NIR.

Furthermore, it is worth noting that a certain interest is being also devoted to both the composition of photoanode (rutile TiO₂ microspheres [18], free standing TiO₂ nanotube arrays [19], ZnO nanowires [20], ZnO/TiO₂ networks [21] have been indeed applied) and materials for counter electrodes [22], dedicated to optimize the overall device composition and fabrication processes.

It should be however pointed out that application of naked QDs is no trivial challenge, due to both chemical stability issues (especially for PbS, which suffers from fast degradation under oxygen ambient and irradiation, when naked and particularly small) and for the varying band gap under size tuning. This latter feature can be highly beneficial once optimized, but could be as well detrimental, if not properly managed, by inhibiting the charge injection from the QD layer and the metal oxide photoanode due to unfavorable band alignment. Both these issues require for careful engineering of the light harvester layer, aimed to equilibrate the goals of efficient charge injection into TiO₂ scaffold, on one hand, and the search for stabilizing the QDs during the device operation, on the other hand.

Recently, we demonstrated how a CdS layer can act as both harvester element and capping agent towards PbS [25], thus allowing a relatively good stabilization of photogenerated current. Herein, we expand the concept, systematically optimizing the fabrication of effective CdS/CdSe and PbS/CdS light harvesters, meeting the needs for efficient light collection in a rather broad range of the solar spectrum as well as a modulation of charge injection through a systematic engineering of harvester band gap (according to the scheme reported in Fig. 1a).

Fig. 1

(a) Scheme of band alignment of the different components of the QDSC. PbS QD bands are qualitatively illustrated according to nanocrystal sizes. (b) Pictures of different QD sensitized photoanodes used in this work (left to right: CdS 4 cycles; CdS/CdSe 5/3; PbS/CdS 2/5; PbS 6 cycles).

CdS, CdSe and PbS were chosen as light harvesters to demonstrate the possibility to effectively engineer the semiconducting sensitizing layer. Application of each of these materials presents both advantages and drawbacks, which are here worth mentioning since, as we will discuss below, they will reflect on device performances.

Although appealing thanks to a good absorption in a broad range (including NIR region), PbS QDs inject with difficulty into TiO₂, because of unfavorable band alignment [28]. Moreover, they easily undergo photocorrosion, thus determining a certain degree of instability in the final solar cell [9]. On the other hand, CdSe QDs show enhanced photogenerated current density [27] together with a good stability in polysulfide electrolyte. However, even neglecting the convenience for deposition under inert atmosphere, due to Se^2- sensitivity to oxygen and moisture, their good activity as photocatalysts has to be mentioned, since it implies a consequent photodegradation over the time, which is of course highly undesirable for photovoltaic devices [29]. CdS QDs present apparently no drawback: they are easy to generate and grow, stable under the device working conditions, and have been demonstrated effective in reducing exciton recombination [31, 32]. However, CdS alone is not able to guarantee satisfying cell performances (photoconversion efficiencies not exceeding 2 % have been recently recorded for CdS alone by applying vertically aligned single crystal TiO₂ nanorods [30]).

As mentioned, in order to overcome the limitations presented by the application of individual semiconductor QDs, composite networks have been recently proposed, such as PbS/CdS [25, 33] and CdS/CdSe [8] The lack of a systematic study analyzing the trend of device performances according to the number of SILAR cycles spurred us in undertaking the present work, in which advantages and drawbacks of application of hybrid QD networks as light harvesters are discussed in terms of device performances.

Experimental section

Results and discussion

Composite PbS/CdS and CdS/CdSe QD networks exhibit contributions to absorbance from each component (namely PbS and CdS, and CdS and CdSe), broadening the absorption spectrum with respect to pure QDs. Incident photon to current efficiency (IPCE) (Fig. 2a) provides for quantitative measure of the efficiency of the process of charge generation, separation, injection and collection. The current density-versus-voltage (JV) curves of selected samples are reported in Figs. 2b and 2c; the quantitative analysis (short circuit photocurrent (J_sc), open circuit photovoltage (V_oc), fill factor (FF) and photoconversion efficiency (PCE)) is collected in Table 1.

Fig. 2

(a) IPCE of selected cells for different kinds of QD. (b) and (c) J–V curves of QDSC sensitized with QDs of various compositions under simulated sunlight (AM 1.5 G, 100 mW cm^–2).

As expected, functional performances delivered by pure PbS-sensitized devices are very poor and show decreasing trend by increasing the number of SILAR cycles (i.e., the PbS sizes). This effect originate from two concurrent phenomena, namely the difficulty of injection determined by the mentioned unfavorable band alignment together with certain instability in polysulfide electrolyte. The first feature is reflected in rather low V_oc values (especially when an excess of PbS is deposited onto TiO₂, see Table 1), which are drastically reduced when increasing the number of SILAR cycles to 6.

On the other hand, satisfying functional parameters are featured by pure CdS-sensitized device (4 SILAR cycles), especially for photogenerated current density, but light collection is possible in a very limited range, centered at around 450 nm (see Fig. 2a). Furthermore, as we will see, J_sc loss in this device under simulated sunlight irradiation is the highest recorded in the analyzed batch.

Addition of CdS to PbS is highly beneficial as for two relevant aspects. CdS QDs are effective in acting as capping agents towards PbS, thus imparting a higher degree of stability to the device. At the same time, photogenerated current is improved, thanks to the extension of light collection in a broader range, as evidenced in the IPCE data reported in Fig. 2a. It is worth noting that there exists an optimal ratio between PbS and CdS: even a little change in the number of SILAR cycles for CdS deposition (4–6 cycles) heavily affects cell performances. The photocurrent density continues increasing under the CdS addition, and the range of light absorption broadens towards the NIR (IPCE at 800 nm around 10 %) with a consequent enhancement of whole IPCE. However, a decrease in V_oc and FF is recorded, which lower the overall performances of the device (Fig. 3).

Fig. 3

Functional properties of PbS/CdS QDSCs as a function of the number of PbS SILAR cycles normalized to the total SILAR cycles (0: pure CdS; 1: pure PbS). (a) V_oc; (b) J_sc; (c) FF; (d) PCE. The dashed vertical line corresponds to the highest PCE (3.0 %).

Further addition of PbS (sample 3/5) is still beneficial in the spectral region 550–800 nm (IPCE at 800 nm around 15 %), but severely affects IPCE in the 450–550 nm spectral region, leading to a moderate increase of PCE with respect to pure CdS (from 1.24 to 1.42 %) and further decreasing V_oc values. The sudden photovoltage decay observed for this sample (Fig. 4) indicates that recombination processes are responsible for the overall decrease of performances. Effect of PbS becomes more evident by analyzing the trend of the device functional parameters versus the number of SILAR cycles applied for sensitization (Fig. 3): fabrication of efficient QDSSCs exploiting PbS QDs mainly deals with finding the best compromise between the V_oc and J_sc determined by this material.

Fig. 4

Transient photovoltage decay of (a) PbS/CdS and (b) CdS/CdSe photoanodes. (c) Electron lifetime for the CdS/CdSe samples, from results in (b).

It should be remarked that the optimum condition for photoconversion efficiency maximization for PbS/CdS network does not correspond to maximum light absorption (see Figs. 2 and 3). Such circumstance is most probably determined by the modification of band alignment of CdS due to addition of PbS, according to the scheme reported in Fig. 1a [34].

Modulating PbS QD sizes implies dealing with a series of physical chemical processes conflicting each other in terms of photoconversion of incident radiation: larger QD size leads to broadened absorption toward IR, but causes a downshift of the conduction band of PbS, inhibiting fast electron injection. In fact, as we discussed above, more than 2 PbS cycles lead to a composite system with limited possibility of fast charge injection into TiO₂. Clear indication comes from the almost linear decrease of V_oc with the increase of PbS (Fig. 3), which testifies the reduced quantum confinement with increase of PbS dimensions and lowering of the conduction band edge of PbS with respect to TiO₂. Maximum light absorption is obtained for the sample 2/6, while the sample 2/5 guarantees the highest photoconversion efficiency (3.0 %). This efficiency is among the highest ever obtained for a PbS/CdS QDSC in the same configuration [25], the “champion” device of similar kind being very recently reported by González-Pedro and co-workers [26]. However, in that case different metal precursors have been used, whose critical role in enhancing functional properties, although highlighted in the paper, had not been clarified. In this frame, very recently, Cao’s group investigated the effect of cationic precursors in SILAR-sensitized QDSSCs, highlighting the role of pH in nanocrystal deposition rate on TiO₂. Nevertheless, other effects ascribable to the metal precursors, such as unexpected red shift of absorbance have not been explained, yet [35].

The application of an engineered CdSe/CdS network results in a remarkable increase of PCE (up to a 3.70 %), although effective light collection does not overcome 650 nm. However, an outstanding increase in IPCE in the 400–650 nm region is observed: the superb CdSe absorption properties, combined with an excellent band alignment with TiO₂, enable charge fast injection. The concomitant occurrence of efficient injection and the rather broad absorption determines a very good IPCE value of 75 % (550 nm) in the best working device (CdS/CdSe 5/5).

Increased number of SILAR cycles above certain threshold does not improve the functional properties of the cell, as testified by the sample 8/5, in which maximum IPCE at 450 nm is around 60 %, as a consequence of poor charge injection and collection, with respect to 5/3 and 5/5 samples.

Transient photovoltage spectroscopy (see Fig. 4) is a powerful tool for analyzing the charge collection and recombination. An almost instantaneous V_oc drop is observed for pure PbS-sensitized devices: unfavorable band alignment results indeed in fast recombination of photogenerated excitons and charge injection inside TiO₂ photoanode is almost completely suppressed. Hybrid PbS/CdS systems show systematic photovoltage decay trend according to the PbS amount: the higher the PbS amount, the faster the decay. The impressively systematic fastening of V_oc decay at increased PbS content is clearly visible in the series 0/4, 2/6, 2/5, 3/5, 6/0. In the mentioned series, as visible in Fig. 4, the pure CdS-sensitized solar cells were found to present the slowest photovoltage decay, testifying again the good band alignment with TiO₂ scaffold allowing for fast electron injection and charge separation.

For the CdS/CdSe system electron lifetime is calculated from transient photovoltage decay (Fig. 4c) by using the following eq. 1 [36, 37]:

Where k_B is the Boltzmann constant, T is the absolute temperature, and e is the elementary charge. In the CdS/CdSe system V_oc decay is much slower, and τ_e can be calculated (see Fig. 4c). τ_e scales with the inverse of QD size, as expected (the total number of SILAR cycles is supposed to lead to larger QDs). In fact, larger QD size results in slower charge injection from the QD to the oxide, affecting electron lifetime, since slow injection can enhance probability of charge recombination [38]. Results in literature refer to colloidal QDs, but they are expected to hold for SILAR, as well.

Stability of short circuit current density was evaluated under pulsed irradiation (for about 1 h) and the best devices sensitized with different QD layers compared (Fig. 5).

Fig. 5

Stability of the short circuit photocurrent density for the best devices for each harvesting layer analysed. J_sc values have been normalized for allowing direct performance comparison. Light is switched on and off at intervals of 240 s.

Current density trend over the time further confirms the beneficial effect of using hybrid QD networks: devices based on this sensitization strategy not only featured improved performances, as discussed, but also present improved photocurrent stability. The highly beneficial role of CdS as capping layer for PbS QDs is further highlighted by the very good stability of the device under irradiation.

It is important noting that the investigation of the crystalline nature of mixed QD networks could be helpful in determining the possible formation of hybrid structures at the interface of the different metal chalcogenides and, consequently, whether these play a role in device functional performances or not. This issue is rarely faced in literature and usually high resolution transmission electron microscopy is applied to single QD sensitized photoanodes (nice analysis, together with discussion of issues related to the measurement, is reported in Ref. [39]). However, very recently the possibility of identifying metal chalcogenide alloys on metal oxide scaffold through Raman analysis has been reported [40]. However, this aspect is beyond the aim of the present study and would deserve a dedicated investigation to be fully clarified.

Conclusions

Despite the relevant number of scientific papers appeared over the last few years on QDSSCs, there is still much room for working in the field. Currently, the most urgent need is probably not related to the race for “champion” devices in terms of photoconversion efficiency: instead, as demonstrated in the present study, more focus should be put in band gap engineering especially devoted to both broad the effective range of light collection and improve device stability over the time.

SILAR technique retains a great potential to fulfil both aims together with a possible sound perspective for future device fabrication scale up; however, recent researches on the exploitation of colloidal QDs as light harvesters indicate that harmonious application in the same device of monodispersed colloidal QDs and polydispersed SILAR-grown QDs could open the path for better devices.

Stability over the time is unfortunately an open issue not fixed yet, although results indicate that a careful materials engineering might face it.

The present study demonstrates that modulation of band gap composition can be easily obtained, thus managing the fabrication of devices with improved performances.

However, efforts from scientific community, able to team different backgrounds up (such as physics, chemistry, engineering) are still needed, aimed to deep the present knowledge and understanding of physic-chemical processes behind the excitonic solar cells.