In this perspective, we take a look back at the successful integration of carbon nanotubes (CNT) into high-efficiency solar cells based on metal-halide perovskites (MHPs). In addition to these successes, we identify critical questions and issues that remain to be addressed for the functionality of CNTs in MHP-based solar cells. Finally, we look forward toward potential future opportunities for CNT/MHP interfaces, in both new types of photovoltaic devices and other emerging optoelectronic applications.
In 2009, Miyasaka and co-workers demonstrated a sensitized solar cell with 3.8% power conversion efficiency that utilized a lead-halide perovskite semiconductor as the photoactive absorber.1 In the decade since that demonstration, lead-halide perovskite solar cell efficiencies have steadily climbed (now to >25% certified efficiency), and metal halide perovskites (MHPs) have generally emerged as a fascinating semiconductor system for both fundamental studies and advanced optoelectronic devices. The success of MHP solar cells can largely be attributed to the combination of a high absorption coefficient, relatively long diffusion lengths, and low exciton binding energies, all of which combine to enable external quantum efficiencies across the absorption spectrum of above 90%.2 These exceptional optoelectronic properties are all the more remarkable given that MHPs are typically processed from solution or deposited via thermal evaporation at relatively low temperatures. As such, MHP solar cells have achieved certified efficiencies rivalling or exceeding those of conventional thin-film semiconductors which tend to require much more sophisticated processing.
The contact materials play an important role in the operation of MHP solar cells, since they introduce the electronic asymmetry in a device underlying the rectifying behavior of a photodiode. In the years leading up to the emergence and proliferation of MHP solar cell research, a number of research groups were exploring alternative contact materials for thin-film solar cells. One such alternative contact was introduced by Wu et al. in 2004—a thin semi-transparent electronically coupled network of single-walled carbon nanotubes (SWCNTs) that struck a delicate balance between high broadband light transmittance and low electrical resistance.3 This demonstration motivated a significant research effort dedicated to integrating SWCNT electrodes (both as transparent electrodes and “back contact” electrodes that need not be transparent) into a variety of thin-film photovoltaic (PV) architectures, including organic solar cells,4–6 cadmium telluride (CdTe),7 copper indium gallium selenide (CIGS),8 and silicon.9 Early research efforts on SWCNT-based contacts set the stage for other novel contact layers such as graphene,10,11 and these materials are still actively explored for solar cells and a myriad of other optoelectronic devices.
Carbon nanotubes (CNTs) were proposed during the nascency of the emerging perovskite PV field as potential p-type contact,12,13 setting off several years of exciting research in this area. Rectification in MHP solar cells is achieved through extraction of electrons and holes on opposite sides of the device at the interfaces with charge-selective contacts [CSCs, Fig. 1(a)]. Selectivity in this context requires rapid transfer of only one charge carrier type, while the other type is blocked. In addition to the essential property of selectivity, in order to achieve high performance devices, CSCs need to have low electrical resistance to avoid charge carrier accumulation, interfacial recombination, and increased series resistance.
In addition to good charge selectivity, contact materials may end up playing an important role in the long-term stability of MHP solar cells. The ease of MHP solution processing arises largely from the ionic nature of the lattice. The flipside of this simple processing is a high vulnerability to degradation through interaction with water, oxygen, heat, and light.14 Furthermore, the low enthalpy of formation and low activation energy for ionic motion allow ionic species to readily move throughout the lattice, in particular, under illumination and/or external bias.15 Ionic motion can have detrimental effects on the device performance, for example, by largely screening the built-in field. Ionic species can also readily diffuse to and react with the contacts at the interfaces,15 in particular, when reactive species such as halide ions are involved.16,17 To enable long-term stability of MHP devices, the materials directly interfacing with the perovskite should consequently be relatively resilient against reacting with the constituents of the perovskite itself. Ideally, the materials should also exhibit barrier type properties against the ingress of moisture and/or oxygen.
As described above, CSCs in MHP solar cells need to combine a diverse set of orthogonal properties in order to allow both high performance and long-term stability. CNT-based contacts have shown encouraging results toward meeting these challenges in a number of different MHP solar cell device architectures, with efficiencies climbing steadily over the past seven years [Fig. 1(b)]. In this perspective, we recount the salient results from these research efforts, paying close attention to studies that provide insights into design rules and mechanistic principles. We then turn to an outlook on some of the remaining questions for CNT-based contacts in MHP-based devices, along with a perspective on the potential role of CNT/MHP interfaces in both new PV-based applications and non-PV optoelectronic devices. We believe that the CNT/MHP interface is particularly versatile, and that future fundamental and applied studies can pave the way for a variety of emerging optoelectronic applications.
II. CARBON NANOTUBES IN PEROVSKITE SOLAR CELLS
In a typical MHP-based solar cell, the perovskite active layer is sandwiched in between two CSCs [Fig. 1(a)]—a hole-selective layer (HSL) and an electron-selective layer (ESL). Early MHP-based solar cells drew inspiration from the dye-sensitized PV community, in which 2,2′,7,7′-tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9′-spirobifluorene (spiro-OMeTAD) and TiO2 had long been popular as HSL and ESL, respectively.18,19 The initial rationale for exploring CNTs as contact material was their intrinsically high charge carrier mobility and long-range transport, which set them apart from conventional p-type materials such as spiro-OMeTAD and poly(bis(4-phenyl) (2,4,6-trimethylphenyl)) amine (PTAA) which both rely on extrinsic doping in N-I-P device configurations. Importantly, dopants such as lithium bis(trifluoromethanesulfonyl)imide (Li-TFSI) have been identified as culprits in rapid device degradation, largely due to their hygroscopicity but also due to Li+ ion ingress.20,21 CNTs have since been explored as CSCs both on the p-type side as well as the n-type side of the perovskite device.22,23 A lot of effort has also gone into exploring the use of CNTs as electrode alternatives, replacing metals such as silver or gold to reduce the potential for metal corrosion and possibly to bring an economic benefit.24
We have explored the use of single-walled carbon nanotube (SWCNT) HSL in various configurations. One of the most interesting findings was that we could use SWCNTs as HSL and combine them with an inert polymer matrix.20 This allows decoupling of the energetic requirements and the stability requirements for the p-type contact. However, in follow-up studies, we showed that SWCNTs as the sole HSL are insufficiently charge-selective for photogenerated holes or cannot generate a strong enough built-in field to prevent interfacial recombination losses.25 This results in a significantly higher photovoltage deficit and, in particular, a much lower steady-state performance (Fig. 2).25 This can be mitigated by p-doping the perovskite surface, thus reducing the minority carrier concentration near the interface.26 However, it remains challenging to achieve performances comparable to those of conventional hole transporters when SWCNTs are used as the sole HSL.
A strategy that appears to be more promising is the combination of CNTs with a secondary hole-transporting material, where the CNTs act as a charge-extraction “highway” from the absorber into the hole transporter.25 The CNTs as an interlayer allow for rapid charge extraction by funneling charges quickly into the secondary material, and recombination at the interface is minimized.27–29 In fact, we could show that using such a SWCNT interlayer with undoped spiro-OMeTAD, which typically suffers from significant losses due to the low charge carrier mobility of spiro-OMeTAD, allowed us to attain close to 21% employing this strategy.30 A generalizable insight we gain from our studies is that the outstanding charge transport properties of SWCNTs can be exploited in conjunction with another hole-transporting material, whose poor charge carrier transport properties can be compensated by those of the SWCNTs.23
While we have focused our studies primarily on devices in the N-I-P configuration, this approach has also been shown to be effective in P-I-N devices, to improve the transport properties of the p-type substrate material such as inorganic NiO using s-SWCNTs and organic binaphthylamine-based hole transporters using MWCNTs.31,32 Despite having a slight p-type character when exposed to ambient oxygen, SWCNTs have also been shown to enhance the charge transport properties of n-type materials such as SnO2 in perovskite solar cells.33,34
The versatility of CNTs as conductive elements has also been demonstrated in hybrid systems where the SWCNTs are directly incorporated into the perovskite absorber with the aim to improve charge carrier transport within the absorber layer.35,36 A thicker absorber layer allows a higher charge carrier yield since a higher fraction of photons will be absorbed. However, when the thickness of the absorber layer exceeds the diffusion length of photogenerated charge carriers, recombination losses result in voltage losses that negate improvements in the photocurrent through a decrease in photovoltage and fill factor. Attempts to maximize the overall device performance, therefore, rely on interventions geared toward improving the charge carrier diffusion length.
Grain boundaries in a polycrystalline film, in particular, tend to impede long-range charge transport.37 Some empirical studies have shown potential techniques to increase the size of the individual grains, for example, by adding ionic liquids.38,39 Another elegant approach is, therefore, the use of SWCNTs as bridging elements, which can thus improve the inter-grain charge carrier transport.35,36 The presence of SWCNTs during the crystallization process can additionally result in a more oriented and larger crystal growth.36 In addition, the amino-functionalized SWCNTs can act as nucleation sites for the polycrystalline perovskite film leading to much larger grain growth, which, in turn, may contribute to better device performances.
Other carbon materials such as fullerenes have been explored in a pseudo-bulk heterojunction with perovskites by blending the fullerenes into the precursor solution or into the anti-solvent.40–42 Such approaches can conceivably passivate the grain boundaries and reduce the migration distance for electrons prior to being extracted by the n-type electrode. Applying this concept to CNTs, they have the advantage of having highly directional charge transport along the CNT axes and could, if aligned appropriately, extract charge carriers rapidly from within the absorber. A few studies have shown that this concept can be employed for photodetectors and solar cells (Fig. 3);43,44 however, those studies still fail to have superior charge extraction compared to conventional planar interfaces.
III. OUTSTANDING QUESTIONS FOR CNT CHARGE EXTRACTION LAYERS IN PEROVSKITES
Despite extensive research over the past six years, there are still several outstanding questions regarding (1) the fundamental mechanism(s) involved in charge extraction and recombination at the CNT/perovskite interface, (2) the degree to which the electronic polydispersity impacts charge extraction at the CNT/perovskite interface, and (3) the ultimate stability of perovskite devices using CNT-based extraction layers or contacts. Here, we provide some perspectives on these important topics.
A. Fundamental mechanisms of charge extraction
The question of fundamental mechanisms is a quite broad one that deserves much longer discussion that we will provide here. Important questions include, but are not limited to, the following. What types of surface rearrangement (e.g., passivation or defect formation) and/or ground-state electronic interactions occur simply by forming the CNT/MHP interface? Do s-SWCNTs or m-SWCNTs have any inherent (dis)advantages for optimizing charge extraction and minimizing interfacial recombination (see discussion below)? Is there a connection between band alignment at the CNT/MHP interface and the kinetics of extraction/recombination? What role do these kinetic and thermodynamic considerations play in determining PV performance metrics and stability?
As the field progresses, answering these questions will require coupling device demonstrations and improvements with spectroscopic, microscopic, structural, and theoretical studies that delve deeply into why particular interfaces facilitate improvements to performance and stability. Below, we briefly discuss the potential (but unclear) role of CNT electronic structure and the underexplored aspect of long-term stability for CNT-containing MHP devices.
B. Role of CNT electronic structure
Most studies utilize as-synthesized SWCNT samples consisting of both s-SWCNTs and m-SWCNTs [Fig. 4(a)] and having a polydisperse mixture of SWCNT diameters with a corresponding polydispersity in frontier orbital energies. As such, the individual (or perhaps synergistic) role(s) of s- and m-SWCNTs in MHP solar cells is unclear. It is not immediately intuitive whether pure s-SWCNTs, pure m-SWCNTs, or a tuned mixture of the two would optimize PV performance. One could envision that a network of pure undoped s-SWCNTs might have the appropriate band alignment for selective extraction of one type of charge, but that it might also be too resistive to effectively shuttle charges away from the interface to reduce interfacial recombination. Improving conductance within the s-SWCNT network would of course be possible through carrier doping,45–47 but this strategy (like the doping strategies used to make spiro-OMeTAD HTLs conductive) could introduce additional ions into a system, where ionic migration already impacts device performance and stability. Additionally, since the electron affinity and ionization potentials of s-SWCNTs can be tuned via diameter, it is feasible that these knobs could be used to tune interfacial band alignment and VOC, but this concept is thus-far completely unexplored.
In contrast to s-SWCNT networks, a pure m-SWCNT network should provide ample conductance, even when undoped, but the high density of accepting states for both electrons and holes over a broad range of energies would not (on paper) appear to facilitate selective extraction of one carrier over the other. Thus, one could envision that pure m-SWCNT networks may serve predominantly to stimulate interfacial recombination. It is tempting to think that a mixture of the two electronic structures may provide the best of both worlds regarding conductance and selectivity, but what ratio is optimal and/or what range of ratios is acceptable?
Surprisingly, the specific role(s) of SWCNT electronic structure in charge extraction has not been systematically studied in detail in MHP solar cells. A number of studies utilize the “CG200” variant of CoMoCAT SWCNTs. Despite these samples having “high metallic tube content” (as reported by the manufacturer), devices made with CG200 composites show no indication of enhanced recombination and in fact were found to produce devices with record-low VOC losses relative to the radiative limit [Fig. 2(c)].25 The precise ratio of metallic SWCNTs in this material is poorly defined, although a couple of studies suggest a range of ca. 40%–48%.48,49 Since the SWCNTs in these HSLs are wrapped with a semiconducting polymer (P3HT) that has shown good selectivity for hole extraction in perovskite solar cells,50 it is unclear if polymer/SWCNT interactions play a role in the selectivity of the m-SWCNT-rich HSL. However, unpublished results from our own studies suggest that insulating polymers such as polyfluorenes (e.g., PFO) can be used to produce efficient devices with CG200 [Fig. 4(b)]. In most cases we have studied thus-far, P3HT wrapping produces devices that have higher stabilized power output, relative to those incorporating insulating PFO polymers, but further systematic studies need to be done. Thus, we suggest that the field would benefit substantially from future studies probing the performance of pure m-SWCNT HSLs, and HSLs spanning a large range of m-SWCNT enrichment. Additionally, it will be illuminating to further explore the impact of the electronic structure of any polymer (either semiconducting or insulating) included in these m-SWCNT enriched HSL composites.
HSLs and interfacial layers enriched in s-SWCNTs have been explored to a limited degree in perovskite solar cells. Figure 4(b) shows previously unpublished data demonstrating that pure undoped s-SWCNT HSLs, with no additional HSL, can produce relatively efficient MAPbI3 solar cells. Dispersion of SG65i CoMoCAT SWCNTs with the insulating polymer PFO (Eg ∼ 3.2 eV) produces samples highly enriched in s-SWCNTs (>99%), with the dominant SWCNT being the (7,5) s-SWCNT.51 Despite the s-SWCNTs being undoped, and with no additional semiconducting polymer present, extracted holes appear to be mobile enough to traverse through the s-SWCNT network to be collected at the back metal contact. Polymer exchange can be performed to replace the PFO wrapping polymer with P3HT,51 but as for metallic-enriched HSLs discussed above, the polymer does not appear to make a significant difference in performance. Again, more systematic studies are warranted.
Time-resolved spectroscopy measurements have demonstrated that pure s-SWCNT networks are indeed highly effective at selectively extracting holes from a prototypical MAPbI3 active layer, and that the resulting charge recombination is exceptionally slow [Figs. 5(a)–5(c)].27,52 PV devices that incorporate these pure s-SWCNTs as interfacial layers between the perovskite active layer and the doped spiro-OMeTAD HSL show improved efficiency and stability for MAPbI3 devices,27 but these extremely thin layers do not need to be conductive since the spiro HTL can shuttle holes to the back contact.
Shapter et al. have recently explored the role of SWCNT electronic structure in the ESL of perovskite solar cells (5D–5F). In this case, they integrated type-separated SWCNTs (HiPCO, d ∼ 1.0 ± 0.2 nm) with varying ratios of s- and m-SWCNTs into the TiO2 ESL of mixed-cation, lead, mixed-halide perovskite solar cells.53 They observed the best performance for TiO2 composite ESLs containing the “natural” ratio of 2:1 semiconducting:metallic SWCNTs (19.35% champion, 18.44 ± 0.89% average efficiency), an improvement over devices containing bare TiO2 ETLs (17.04% champion, 16.64 ± 0.60% average efficiency). The authors suggest that (1) leakage current paths can be suppressed when s-SWCNTs are integrated into the TiO2 electrode; (2) m-SWCNTs reduce the electrode impedance but can enhance back recombination of carriers at the interface and (3) a combination of the two types of SWCNTs takes advantage of the beneficial properties of each SWCNT type. Interestingly, devices containing pure s-SWCNTs in the TiO2 composite showed faster degradation, whereas all other SWCNT-containing devices were more stable than those containing bare TiO2 electrodes.
These encouraging initial results suggest that the SWCNT electronic structure may impact the performance of charge extraction electrodes and should continue to be explored in a systematic fashion. Perhaps the most encouraging conclusions from these results is that the SWCNT electronic structure does not dramatically affect the PV device performance and devices prepared from as-synthesized (unseparated) commercially available SWCNTs appear to work quite well. From a cost and manufacturing perspective, this means that expensive separation and enrichment procedures would likely be unnecessary.
Stability remains an important question for MHP-based solar cells, and several studies have explored the short-term stability of CNT-containing MHP devices. The ultimate question for CNT-based electrodes will be whether they can in fact yield perovskite devices that maintain high efficiencies over a long period of time without a large performance loss. To a large extent, previous studies have relied on proxy measurements such as improved thermal stability or reduced performance loss under dark storage conditions in order to make improvements in stability when employing CNTs. Stability tests for perovskite solar cells are slowly shifting to a more rigorous testing paradigm.54,55 According to which, more and more studies demonstrate stable performance under full illumination and at elevated temperatures tracking the maximum power-output, thus being more representative of real-world conditions.38 Those types of studies are still lacking for CNT-based electrodes.
A particular concern for MHPs is the mobility of the various ions that make up the material. In particular, mobile halides and their vacancy counterparts have been identified to lead both to instabilities of the absorber but also the contacts due to corrosive reactions.14,17 Especially the detrimental interactions between egressing halide and penetrating metal ions have been shown to impact the long-term device stability with various metal electrodes.17,56,57 For the development of MHPs as a viable PV technology, it is, therefore, a pressing issue to devise strategies which can minimize these types of ionic interactions in order to push the operational lifetime of MHP devices toward 20 years and beyond. One such approach uses the interfacial layer with the MHP absorber as a blocking layer for ions. In this context CNTs can enable other materials such as in-filled inert polymers as the blocking layer, while the CNTs facilitate the charge transport [Figs. 6(a)],20 or CNTs can also act as an ion blocking layer in their own right.58 In fact, the layer of CNTs in this device stack appears to block both the migration of iodide to the metal electrode, but also the migration of metal ions into the MHP absorber thus significantly prolonging the device lifetime.58
Another approach to tackle the stability issues of metal electrodes in perovskite solar cells is to replace the metal altogether. Carbon-based electrodes have emerged as a possible alternative with the prospect of being more resilient towards corrosion.59 As highly conductive carbon materials, CNTs have been explored as interlayers between the perovskite and the carbon electrode with the aim to improve the charge extraction to the electrode.60 Replacing the electrode completely with CNTs has remained challenging for a long time since the sheet resistance of untreated CNTs tends to be higher than that of metal electrodes.46,61 Progress toward these goals has been achieved through strategies aimed at doping the CNT film prior to its deposition,62 an approach that recently resulted in a CNT-based electrode outperforming a conventional silver electrode.24
Another area for additional research involves the stability of flexible MHP-based solar cells. While these types of devices can enable the proliferation of PV into a huge swath of everyday mobile applications that are not feasible for traditional PV, such devices should also be able to undergo frequent and/or repetitive deformations without efficiency losses. The low-temperature solution processability of MHP solar cells sets this technology apart, because it allows for roll-to-roll processing on cheap, flexible substrates which can give perovskite photovoltaics a significant advantage over most other PV technologies in terms of scalability and cost.63,64 While there remain open questions with regards to the mechanical stability of the MHP absorber itself, forward-looking device approaches need to also consider the mechanical resilience of the contacts since the limitations of a full flexible device stack are defined by the most brittle layer. A unique aspect of CNTs is that their conjugated double bonds confer both excellent charge transport properties as well as exceptional mechanical resilience.65 This makes CNTs particularly attractive as a contact material in flexible device stacks.66,67 Several studies have shown the successful integration of CNT-electrodes in flexible perovskite solar cells in both conventional N-I-P and P-I-N architectures [Figs. 6(b) and 6(c)].33,68,69 In an interesting study exploiting the flexible properties of CNTs, researchers fabricated perovskite solar cells as flexible fibers [Fig. 6(d)].70
A final stability consideration for CNTs in MHP solar cells involves the CNTs themselves. While CNTs are generally quite stable on their own, there are still only limited studies of the long-term stability of CNTs in functioning devices under operating (and sometimes extreme) conditions. Oxidative environments have been shown to convert sp2 SWCNT bonds to sp3 bonds, forming covalently bound oxygenic species (e.g., epoxides, ethers, etc.).72 These sp3 bonds disrupt the pi electron network of the SWCNTs, diminishing conductivity. As with other SWCNT functionalization routes, sidewall reaction is enhanced for SWCNTs exposed to light above the optical bandgap and elevated temperatures. While most of these experiments utilize relatively extreme oxidative environments (e.g., ozone, energetic oxide deposition) to intentionally create the sp3 defects, it is currently unclear the degree to which ambient oxygen and/or water may induce such transformations for routine long-term operation of solar cells containing CNTs. To this extent, transport layers containing DWCNTs or MWCNTs may prove beneficial, since even if the outer nanotube wall reacts the inner wall(s) may remain intact and conductive.73
IV. AN OUTLOOK ON FUTURE OPPORTUNITIES FOR CNT/PEROVSKITE INTERFACES
While CNT-based charge transport layers have clearly shown good potential for serving various roles in MHP solar cells, there are still untapped opportunities for CNTs in MHP-based optoelectronic devices. Here, we discuss several potential MHP-based devices and architectures that may capitalize upon the unique properties of CNTs.
A. Tandem solar cells
Tandem devices represent a viable strategy to increase solar cell efficiencies, while keeping additional costs low enough for ultimately reducing the levelized cost ($/kWh) of delivered solar electricity.74 There are numerous potential options for tandem solar cells, both in terms of the material choices for top and bottom cells as well as device interconnection (i.e., two-, three-, or four-terminal).75 Perovskite-on-perovskite tandems are a technology that is especially promising for fully flexible devices that in theory could be fully solution-processed. Recent studies used a FA0.75Cs0.25Sn0.5Pb0.5I3 low-bandgap (1.27 eV) cell with a FA0.6Cs0.3DMA0.1PbI2.4Br0.6 wide-bandgap (1.7 eV) cell to achieve 23.0% stabilized power output (SPO) for rigid devices and 21.3% SPO for flexible devices (Fig. 7).76 CNT-based layers or composites could likely serve as suitable recombination layers in such devices (see the PEDOT:PSS/AZO/IZO layer in Fig. 7).
Perovskite-on-silicon tandems have received increasing attention as a viable device configuration that takes advantage of both the established silicon PV infrastructure for an efficient red-absorbing Si bottom cell and the low capital expenditure associated with a solution-processed blue-absorbing perovskite top cell.77,78 Interestingly, besides the demonstrated successes in perovskite solar cells, another PV technology in which CNTs have shown exceptional promise is in silicon heterojunction devices. Most recently, simply fabricated c-Si heterojunction devices based on Phosphorous-doped n-type crystalline silicon wafers with a CNT/Nafion hole extraction layer have achieved efficiencies >21%.9
The good charge extraction characteristics of CNTs in both perovskite and Si-based solar cells suggest that there may be a variety of ways in which CNTs could serve in interlayers between the two in tandem cells. A recent advance in perovskite-on-Si tandems utilized a fully textured Si heterojunction bottom cell and a thick (micrometer) Cs0.05MA0.15FA0.8PbI2.25Br0.75 perovskite top cell (Eg = 1.68 eV).78 The perovskite NiOx HTL sits atop the silicon bottom cell. The authors demonstrate that a key advantage of the structured Si surface is to increase the depletion width and the contribution of drift to the photocurrent such that charges can be efficiently extracted from the bulk of the thick perovskite layer. One could envision that a CNT-based HTL could further assist in optimizing charge extraction in such a device architecture, since (if deposited appropriately) the high aspect ratio CNTs could penetrate substantially into the perovskite active layer. In such a geometry, one of course has to consider the optical losses associated with the CNT layer, but our demonstration of thin monochiral SWCNT interfacial layers with exceptionally good charge extraction suggests that optical losses could be kept to a minimum for thin layers since the dominant S11 absorption (1000 nm or 1050 nm) is very narrow. The conformal nature of CNT HTLs also imply they could be adapted to a broad array of Si bottom cell textures.
B. Perovskite nanocrystal solar cells
Metal-halide perovskites can also be synthesized as colloidally stabilized nanocrystals (NCs), also referred to as quantum dots (QDs), when the degree of spatial confinement approaches the excitonic Bohr diameter of the material in question. MHP NCs have the advantage of being fully formed into the perovskite crystal phase in solution79 with crystallite diameters that are controlled by the synthesis conditions and provide size-tunable optical bandgaps. As such, the NCs can be fabricated into electronically coupled NC arrays with widely tunable absorption properties, even having novel properties such as graded bandgaps within the NC array that are not possible for traditional perovskite layers80 [since sequential deposition of bulk perovskites with different bandgaps would typically dissolve the lower layer(s)]. Solar cells based on perovskite NCs have steadily increased in efficiency over the past ten years, most recently reaching certified record efficiencies of 16.6% this year.81
CNT-based charge extraction layers could potentially be quite beneficial in perovskite NC solar cells, although they are currently unexplored for this device architecture. The high aspect ratio of CNTs could be particularly advantageous for improving charge extraction in NC-based devices since charge transport in NC arrays is less efficient that in bulk perovskite films. The CNTs could effectively penetrate relatively deep into the NC array, creating a morphology somewhat akin to the bulk heterojunction morphology used ubiquitously for organic photovoltaics. Since the NCs are already crystallized in the perovskite structure in solution, co-deposition of CNT/NC solutions could be used as a tool to create blended films in a stepwise fashion without concern that the CNT inclusions may hamper the crystallization dynamics of the perovskite layer. Indeed, one of the biggest limitations for NC layers is the relatively poor long-range charge transport. Hence, the performance of NC-based solar cells is often limited by the trade-off between maximizing absorption and being charge-transport limited. The integration of CNTs into a NC/CNT blend structure could significantly improve the transport properties of such a layer and allow NC-based solar cells to maximize their photocurrent yield.
Recent results demonstrate that SWCNT-based hole extraction layers can improve the performance of non-perovskite (e.g., PbS) based NC solar cells.82 As such, we believe that such interfaces may be quite promising for the development of efficient and stable perovskite NC solar cells in the years to come.
C. Additional opto-electronic devices utilizing CNT/perovskite interfaces
We close by considering a number of additional perovskite optoelectronic devices, beyond traditional solar cells, that may benefit from CNT-based interfacial materials.
1. Light-emitting diodes
MHPs have emerged as exciting semiconductors for light-emitting applications. MHP emission can be tuned across the entire visible range, even in the technologically difficult green region of the spectrum, by manipulating perovskite stoichiometry and/or dimensionality. In many systems, self-trapped excitons produce broadband emission that could potentially be used to enable white-light LEDs.83 The demonstrated successes of CNT charge extraction layers in perovskite solar cells suggest that they could also be beneficial for charge injection in perovskite LEDs.
A few recent studies have explored this emerging opportunity. Bade et al. fabricated flexible MAPbBr3 perovskite LEDs on CNT electrodes with green emission (ca. 550 nm) that could be bent to a radius of 5 mm.84 Andricevic et al. grew MAPbBr3 single crystals between two vertically aligned CNT forests to produce a green LED with an average luminance of 60 cd/m2 at room temperature.85 The authors suggest that the device operates in similar fashion to a light-emitting electrochemical cell (LEC), since it capitalizes upon charged ion drift in the applied electric field. Most recently, Jamali et al. used a solution-based layer-by-layer coating technique to produce green-emitting (ca. 530–550 nm) perovskite-coated carbon nanotube light-emitting fibers.86 The coaxial fiber consisted of a solution-spun CNT fiber coated first with a composited containing the MAPbBr3 perovskite in polyethylene oxide (PEO) and finished with a silver nanowire conducting film.
The excellent performance of CNTs in MHP solar cells suggests that similar success may be achieved in photodetectors. CNTs have been integrated into MHP-based photodetectors in many recent studies that utilize a variety of different MHP stoichiometries and dimensionalities.43,87–90 As with solar cells, CNT-based detectors open up possibilities for unique flexible architectures88 and can also enable detection in the near-infrared.90 The study of CNT/MHP interfaces in photodetectors is relatively young but is an intriguing direction with a lot of potential.
3. Adaptive/switching optoelectronic devices
A final area for future exploration of CNT/MHP interfaces is the broad field of adaptive optoelectronic devices. Such devices respond to external stimuli (light, temperature, voltage, etc.) in such a way that their optical and/or electrical behavior reflects, or adapts to, the external environment. Wheeler et al. recently demonstrated an adaptive MHP-based solar window that can be switched between a “bleached” window state and a “colored” PV state.91,92 As shown in Fig. 8, solar photothermal heating evolves methylamine from an optically transmissive MHP-methylamine complex to produce the black MAPbI3 photovoltaic absorber, whereas cooling allows methylamine to intercalate back into the device to form the complex. Such “smart window” technologies have the potential to reduce energy consumption and enhance comfort in buidings. A SWCNT interfacial layer played a key role in the first incarnation of this device,91 since the hole extraction layer must be (1) highly conducting, (2) selective for hole extraction, (3) transmissive in the visible for window functionality, and (4) porous to enable methylamine to escape and re-intercalate into the device. Meeting these requirements simultaneously is a challenging task that is not possible for most charge extraction layers.
Another example of adaptive/switching behavior can be found in the field of neuromorphic information processing, where (opto)electronic materials respond to external stimuli by changing their conductance (conductance switching), often in an analog fashion where the conductance at any given time reflects the precise history of excitation. Such “artificial synapses” can be used in materials-based strategies to emulate the signal processing and memory behavior of synapses and neurons in the brain. Looking forward, we note that opportunities exist for CNT/MHP interfaces in electrical and optical memories and artificial synapses.93,94
In this perspective, we highlight the progress made toward incorporating multi-functional CNT-based layers into high-efficiency MHP-based solar cells. We also discuss a number of outstanding questions we hope that the community can address in future studies for understanding the performance and stability of CNT/MHP interfaces. The successes realized in MHP-based solar cells suggest that CNT/MHP interfaces can be used to enable a variety of unique functionalities to emerging MHP-based optoelectronic devices, including tandem solar cells, light-emitting diodes, photodetectors, and adaptive/switching technologies. We believe that these technologies have the potential to broadly impact energy harvesting, storage, and efficiency in the years to come, and we are excited to follow (and contribute to) the progress in these areas.
J.L.B. gratefully acknowledges funding from the Center for Hybrid Organic Inorganic Semiconductors for Energy (CHOISE), an Energy Frontier Research Center funded by the Office of Basic Energy Sciences, Office of Science, within the U.S. Department of Energy (DOE) through Contract No. DE-AC36-08GO28308. S.N.H. was supported by the Director's Fellowship program of the National Renewable Energy Laboratory. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes.
The data that support the findings of this study are available from the corresponding author upon reasonable request.