Synthetic Metals 222 (2016) 17–22

Contents lists available at ScienceDirect

Synthetic Metals
journal homepage: www.elsevier.com/locate/synmet

Carbon quantum dots as new hole transport material for perovskite

solar cells
Sofia Pauloa,b , Georgiana Stoicaa , Werther Cambaraua , Eugenia Martinez-Ferrerob,** ,
Emilio Palomaresa,c,*
a
Institute of Chemical Research of Catalonia (ICIQ), The Barcelona Institute of Science and Technology (BIST), Avinguda del Països Catalans 16, 43007
Tarragona, Spain
b
Centre Fundació Eurecat, Av. d’Ernest Lluch 36, Parc Científic i de la Innovació TecnoCampus, E-08302, Mataró, Barcelona, Spain
c
Catalan Institution for Research and Advanced Studies (ICREA), Passeig de Lluis Companys 23, 08010 Barcelona, Spain

A R T I C L E I N F O A B S T R A C T

Article history:
Received 3 November 2015 The use of carbon quantum dots (CQDs) in photovoltaic devices has not yet been fully explored. Herein,
Received in revised form 20 April 2016 we report the use of CQDs in MAPI (methyl ammonium lead iodide) perovskite solar cells as a first
Accepted 24 April 2016 attempt to implement this material in photovoltaic applications beyond its use as transparent conductive
Available online 19 May 2016 electrode.
ã 2016 Elsevier B.V. All rights reserved.
Keywords:
Carbon quantum dots
Perovskite
Hole transport materials
Photovoltaic devices
Hybrid solar cells

1. Introduction researchers have proposed their use in solar energy application

During the last decade in the areas of organic electronics and
biotechnology, two new carbon based materials, namely carbon
quantum dots (CQDs) and graphene quantum dots (GQDs), have
grabbed the attention of numerous research groups thanks to their
potential to be used either in optoelectronic devices or bio-
imaging, respectively [1–3].
In biotechnology, CQDs have revealed interesting properties
like water solubility, low toxicity (especially compared with their
inorganic counterpart QDs like PbS or CdSe), stability against
photobleaching and better biocompatibility [4].
For optoelectronic applications, particularly since 2010, when
the Nobel Prize in physics was awarded to A. Geim and K.
Novoselov for their work on the 2D material graphene, many
Abbreviations: CQD, carbon quantum dot; HTM, hole transport material; MAPI,
methylammonium lead iodide perovskite; GQD, graphene quantum dots; ETM,
electron transport material; FTO, fluorine tin oxide; PL, photoluminescence; HOMO,
highest occupied molecular orbital; LUMO, lowest unoccupied molecular orbital.
* Corresponding author at: Institute of Chemical Research of Catalonia (ICIQ), The
Barcelona Institute of Science and Technology (BIST), Avinguda del Països Catalans
16, 43007 Tarragona, Spain.
** Corresponding author.
E-mail addresses: eugenia.martinez@eurecat.org (E. Martinez-Ferrero),
epalomares@iciq.es (E. Palomares).
such as hybrid and organic solar cells, light emitting diodes or
photodetectors [3].
In this work, we study the possibility to employ a solution
processed layer of CQDs as HTM in perovskite solar cells. This type
of solar cell is nowadays trending topic in the photovoltaic research
world mostly due to the outstanding power conversion efficiencies
(15–20%) reached in just a few years from its appearance [5].
In these years, a lot of effort has been put to improve the
stability and performance of the devices either by developing
alternative ways to form the perovskite layer (absorber) [6–10] or
employing new materials to facilitate charge transport through the
device (selective contacts) [11–14]. The latter pathway is usually
focused on testing and development of new HTMs, like polymers or
small molecules, to be deposited via solution processes on top of
the perovskite layer, being the titanium dioxide (TiO2) the most
employed material as an electron transport material (ETM). This
research activity could be conceived as an open race to find
competitors for spiro-OMeTAD (which is right now the most used
HTM in this type of cells) that can be synthesized in an easier way
(fewer synthetic steps), in large scale, that do not need dopants or
oxidants and, above all, that could lead to an increase in stability
and performance of the solar cell.
In this paper, we propose the use of solution processed CQDs as
HTM in MAPI solar cells. In Section 2 we will describe the synthesis

http://dx.doi.org/10.1016/j.synthmet.2016.04.025
0379-6779/ã 2016 Elsevier B.V. All rights reserved.
18 S. Paulo et al. / Synthetic Metals 222 (2016) 17–22
(2.1) and the device fabrication (2.2) and in Section 3 we will show
the results obtained in this study.
2. Experimental
2.1. Synthesis of carbon quantum dots
The synthesis was carried out via a polymerization–carboniza-
tion hydrothermal approach [15]. 1.0507 g of citric acid and 0.541 g
of p-phenylenediamine were well dissolved in 10 mL of deionized
water. The solution was transferred to a poly (tetrafluoroethylene)
(Teflon)-lined autoclave (120 mL) and heated at 200
C for 5 h, with
a heating ramp of 5
C/min. After the reaction, a brown-black
solution was obtained. Then, the solution was centrifuged thrice at
4500 rpm and the CQDs were dispersed in deionized water
(CQDaq).
2.2. Device fabrication
Fluorine doped Tin Oxide (FTO) coated glasses (resistance 8 V/
cm2) were etched using Zn powder (Alfa Aesar 98%) and a 2 M
solution of HCl according to the desired pattern. Afterwards, the
substrates were cleaned for 15 min in an ultrasound bath with
deionized water with Hellmax soap, with deionized water and
finally with ethanol. The substrates were dried and an UV/Ozone
treatment was performed for 20 min.
0.65 mL of Ti (IV) isopropoxide (Sigma–Aldrich 97%) and
0.38 mL of acetylacetone (Sigma–Aldrich) were mixed in 5 mL of
ethanol. This solution was deposited by spin coating at 3000 rpm
for 60 s on top of the FTO. Substrates were calcined at 500
C for
30 min to obtain the TiO2 dense layer (d-TiO2) having a thickness of
around 50 nm. Afterwards, titanium oxide-coated substrates were
immersed in a 40 mM TiCl4 solution at 70
C for 30 min. Then, the
substrates were cleaned with water and ethanol and heated at
390
C for 20 min.
A mixture of TiO2 paste (18 NR-T Dyesol) and ethanol 2:7 (w/w)
was spun at 5000 rpm for 30 s. The substrates were heated at
500
C for 30 min to form a mesoporous layer (mp-TiO2) of around
350 nm.
The methylammonium iodide was synthesized as described
previously [16]. The MAPI solution consists of a mixture of 3:1
(molar ratio) of methylammonium iodide and PbCl2 (Sigma–
Aldrich 98%) in DMF [10]. 80 mL of this solution were spun above
the mp-TiO2 layer at 2000 rpm for 60 s. The as deposited substrates
were heated at 100
C for 1 h.
In order to avoid perovskite dissolution, the water dispersed
CQDs were redissolved in isopropanol (CQDiso) to obtain a 2.5 mg/
mL solution. 80 mL of CQDiso were spun at 1000 rpm for 90 s.
72.3 mg of spiro-OMeTAD (1-Material) were dissolved in 1 mL
chlorobenzene with the addition of 28.8 mL tert-butylpyridin and
17.5 mL of lithium bis-trifluoromethanesulfonimide solution in
acetonitrile (520 mg/mL). The resulting solution was spun (80 mL)
onto the perovskite at 2000 rpm for 60 s.
Fig. 1. Schematic representation of CQDaq synthesis. Picture of the CQDaq under 360 nm UV lamp.
Finally, a layer of 80 nm of gold was deposited as the anode by
thermal evaporation at a pressure not higher than 1 106 mbar.
All the steps for MAPI and HTM deposition were carried out
inside a nitrogen filled glove box to avoid the excess of humidity
([O2] < 100 ppm and [H2O] < 0.1 ppm).
2.3. Characterization
UV–vis and fluorescence spectra were recorded, using a 1 cm
path length quartz cell in a Shimadzu UV spectrophotometer 1700
and a Fluorolog Horiba Jobin Ivon spectrophotometer, respectively.
Time-Correlated Single Photon Counting (TCSPC) measurements
were carried out on an Edinburgh Instruments LifeSpec-II
spectrometer equipped with the appropriated PMT detector,
double subtractive monochromator and a set of picosecond pulsed
diode lasers sources. Cyclic voltammetry (CV) experiments were
performed with an Autolab PGSTAT 30 electrochemical analyzer at
room temperature with a three-electrode configuration in a
dimethylformamide solution containing the sample (typically
about 1 mM) and the supporting electrolyte. A platinum (ø 1 mm)
disc served as the working electrode, a platinum wire (ø 1 mm) and
a commercial Ag/AgCl electrode being the counter and the
reference electrodes respectively. 0.1 M tetrabutylammonium
hexafluorophosphate (NBu4PF6) was employed as the supporting
electrolyte. The solutions were stirred and deaerated by bubbling
argon for a few minutes prior to each voltammetric measurement.
The scan rate was 0.01 V/s.
The J–V characteristics of the devices were measured using a
Sun 2000 Solar Simulator (150 W, ABET Technologies). The
illumination intensity was measured to be 100 mW/cm2 (equiva-
lent to 1 Sun) with a calibrated silicon photodiode (NREL). The
applied potential and cell current were measured with a Keithley
2400 digital source meter. Environmental scanning electron
microscopy (ESEM) and Energy Dispersive X-Ray Spectroscopy
(EDX) were carried out at low voltage (20 keV) in a FEI Quanta 600
FEG microscope equipped with an EDX detector from Oxford
Instruments. ImageJ software was used for surface coverage
quantitative image analysis.
3. Results and discussion
3.1. Carbon quantum dots (CQDs)
CQDs were synthesized following a one-step hydrothermal
treatment of citric acid, as carbon source, and p-phenylenediamine
as the surface passivation agent, as depicted in Fig. 1. According to
the synthesis mechanism [17], the CQDs are formed through four
consecutive stages: dehydration and polymerization (both facili-
tated by the carboxyl groups of the citric acid), carbonization, and
passivation, respectively. Initially, the molecules assemble as a
result of hydrogen bonding. Then, during heating (dehydration),
polymerization occurs, leading to a short single burst of nucleation.
 S. Paulo et al. / Synthetic Metals 222 (2016) 17–22 19

Fig. 2. Absorbance (a) and emission (b) spectra of CQDaq (solid line) and CQDiso (dashed line), respectively. Inset photo in Fig. 2b represents the CQDiso under 360 nm UV
lamp.

The resulting nuclei then grow by the diffusion of solutes toward

the particles surfaces.
The amine molecules play a dual role as n-doping precursors
and as passivating agent, which finally is translated into improved
photoluminescence. Heteroatom doping (N, F, S, P, Gd) is a
commonly used approach to fine-tune or obtain new photo-
luminescence (PL) and other physicochemical properties of
photoluminescent materials [18]. Furthermore, co-doping strate-
gies by combining Mg/N, P/N, and S/N were found to dramatically
improve the photoluminescent emission, in addition to lumines-
cence stability, high solubility, low toxicity, and excellent
biocompatibility [18]. The UV–vis spectrum of the CQDaq (solid
line in Fig. 2a) shows an absorption peak around 380 nm, very well
defined for the CQDiso nanoparticles (dashed line in Fig. 2a). As
reported previously [19], the peak located at 380 nm is ascribed to
the presence of the C¼O bonds and their n–p* transition, as a
consequence of amides formed between  NH2 and COOH,
while the broad tail extending over the visible part of the spectrum
Fig. 3. PL lifetime decay of CQDs in water (solid line) and isopropanol (dashed line).
likely originates from other functionalized surface groups of the lexc = 405 nm, lems = 452 and 514 nm respectively. Solid lines represent the
CQDs. biexponential fit of the decays.
Next, the corresponding PL spectra of both samples were
recorded upon excitation at 380 nm. CQDaq displayed an emission
centered at 515 nm (solid line in Fig. 2b). Additionally, a greenish
The decay curves can be fitted to a biexponential model
color PL emission was observed upon excitation of the CQDaq at
described in Eq. (1).
360 nm with an external UV lamp, as displayed in Fig. 1. On the
other hand, the PL emission of CQDiso experienced a blue shift of F(t) = B + B1exp(t/t1) + B2exp(t/t2) (1)
60 nm with a maximum at 455 nm displaying a bright blue color
where t1 and t2 represent the time constants and B1 and B2
when excited with the UV lamp (360 nm), as indicated by the inset
represent the amplitudes of the components, respectively. The
photo in Fig. 2b. These results indicate that solvents with different
average lifetime was calculated by the expression in Eq. (2).
relative polarities and dielectric constants have different influence
on the optical properties of the CQDs despite both of the solvents 4t = (B1t12 + B2t22)/(B1t1 + B2t2) (2)
being polar protic. In other words, a less polar solvent like the
The fitting parameters B1, t1, B2, t2 and t are summarized in Table 1.
isopropanol (dielectric constant of 18 and relative polarity of 0.546,
The average life time was calculated to be 3.61 ns for the CQDs
respectively [20]) will induce a blue shift of the CQDs emission
aqueous solution and 12.38 ns for the CQDs in isopropanol pointing
with respect to water (dielectric constant of 80 and relative
to the influence of the solvent on the lifetime, in line with the PL
polarity of 1, respectively). This solvent-dependent PL emissions
fluorescence behavior. This effect is more pronounced on the slow
behavior is compatible with the previously reported study by Zhu
component, while the fast component, attributed to the radiative
et al. [21] when investigating GQDs. The authors registered a PL red
recombination of the excitons responsible for the fluorescence,
shift from 515 nm to 575 nm when dissolving the carbon nano-
particles in solvents with different polarities (THF, acetone, DMF,
Table 1
water), as a consequence of the solvent attachment or different Time constants t1 and t2, components B1 and B2, and average lifetime t of CQDaq
emissive traps on the surface of quantum dots. and CQDiso, respectively.
Fig. 3 represents the luminescence decay profile of both the
Sample B1 (%) t1 (ns) B2 (%) t2 (ns) t (ns)
aqueous and isopropanol solutions of CQDs, respectively. This
CQDaq 54.62 1.60 45.38 4.47 3.61
analysis provides average radiative lifetimes those contain
CQDiso 29.05 3.32 70.95 13.31 12.38
contributions from both excitonic and trap states.
20 S. Paulo et al. / Synthetic Metals 222 (2016) 17–22

Fig. 4. UV–vis and emission spectra of CQDiso (a) and cyclic voltammetry (b) of CQDs in DMF.

Fig. 5. Depiction of the device (left) and energy levels of the materials employed (right) with the device architecture FTO/d-TiO2/mp-TiO2/MAPI/HTM/Au, where HTM is either
CQDs or spiro-OMeTAD.

hardly changed. Similar results have been reported for mercapto-

propionic acid-capped CdTe quantum dots which displayed a
significantly longer lifetime in methanol than in water [22]. These
results drive us to the conclusion that the solvents affect not only
the PL emission but also the lifetime of the QD nanoparticles, with
emphasis on the polarity and the interaction between the
functional groups on the nanoparticle surface and the solvent,
respectively. The electrochemical cyclic voltammetry was per-
formed for determining the highest occupied molecular orbital
(HOMO) and the lowest unoccupied molecular orbital (LUMO)
levels of the CQDs. Fig. 4b shows the cyclic voltammogram of CQDs
solution in DMF.
The CQDs oxidation potential was determined as + 0.78 eV. The
HOMO of the CQDs was thus calculated using the Eq. (3).
EHOMO = (4.88  E1/2,Fc,Fc+ + EOX) (3)
where EOX is the oxidation potential of the CQDs and E1/2,Fc,Fc+ is the
half-wave potential of Fc/Fc+ estimated in 0.54 eV. Thus, the HOMO Fig. 6. Current density-voltage characteristics of the MAPI devices with and
has a value of 5.12 eV. The LUMO (2.07 eV) was then calculated without the HTMs studied in this work. The measurements have been performed in
by the equation 4. reverse scan (from higher to lower voltages).

5 ELUMO = EHOMO + E0–0 (4)

Table 2
where E0–0 represent the 0–0 energy, which is defined as the Device performance of the MAPI solar cells shown in Fig. 6.
lowest energy transition and that we can estimate from UV–vis and
HTM JSC (mA/cm2) VOC (V) FF (%) PCE (%)
fluorescence emission spectra. As shown in Fig. 4a, we obtained a
value of 3.05 eV for E0–0 corresponding to 407 nm. The HOMO CQDs 7.83 0.515 74 3.00
OMeTAD 17.08 0.720 66 8.06
energy value is above the MAPI perovskite valence band (VB)
NO HTM 7.58 0.223 42 0.71
energy (-5.5 eV), which ensures efficient hole transfer from the
 S. Paulo et al. / Synthetic Metals 222 (2016) 17–22
Fig. 7. Top view ESEM images of CQDs and spiro-OMeTAD (reference) devices.
perovskite to CQDs (Fig. 6). For comparison purposes, in Fig. 5, the
HOMO LUMO levels of spiro-OMeTAD are also shown.
Photovoltaic devices were fabricated to evaluate the perfor-
mance of the CQDs as HTM in comparison with spiro-OMeTAD. The
J–V characteristics shown in Fig. 6, demonstrate the evidence that
the CQDs layer introduces a benefit with respect to the absence of
an HTM in the device, although the reference cell with spiro-
OMeTAD performs much better in terms of short circuit current
density (JSC), open-circuit voltage (VOC) and, thus, in power
conversion efficiency (PCE). The increase in VOC and FF upon
insertion of a CQD layer suggests that the latter fairly improves the
contact between the perovskite and the gold electrode.
Numeric values for reverse voltage scan (see SI for both forward
and reverse) are summarized in Table 2
Fig. 7 shows representative top view ESEM images of a device
fabricated with CQDs, together with the spiro-OMeTAD reference
solar cell. In both cases, by simple optical evaluation of the images,
it can be observed that their surface was not fully covered by the
perovskite layer. Indeed, an image treatment with the ImageJ
software revealed poor perovskite coverage over the mp-TiO2
surface, i.e. around 60% for the spiro-OMeTAD cell and 50% for the
CQDs device, respectively (Figs. SI1a, SI1b). The relative poor
coverage of the mp-TiO2 could be responsible for the low
efficiencies of the devices compared to ones reported in the
literature as indicated previously by other groups [9]. The fact that
the JSC of the device does not increase with the introduction of a
CQD layer could be due to different factors. Even if the SEM images
do not give us a proof of evidence, we consider that the CQD layer is
not homogeneous since single dots tend to aggregate forming
island-like domains. This configuration could lead to a bad charge
transfer between the perovskite and the top contact. Another
possible reason for the low photocurrent could be attributed to the
type of ligand employed for the CQD (p-phenylenediamine) which
could be responsible of a lack of charge mobility.
The spectrum profile of the elemental mapping by EDX clearly
shows the presence of Pb and I, while Cl was not detected in either
of the devices (Fig. SI2). The quantitative analysis revealed a molar
ratio of Pb/I = 3.45 for the CQDs and Pb/I = 3.7 for the spiro-
OMeTAD, respectively, which is slightly higher than the theoretical
ratio of the MAPI solar cells (Pb/I = 3). These results demonstrate
that the perovskite was successfully formed during the synthesis
step.
6. Conclusions
In conclusion, the synthesis and optical and electrochemical
characterization of CQDs have been extensively described. The
21
cyclic voltammetry measurements revealed that the HOMO and
LUMO energy values of the CQDs are appropriate to ensure both
efficient hole transfer and electron blocking capability from the
perovskite to CQDs. Preliminary photovoltaic devices based on
CQDs as HTM have been fabricated for the first time and a power
conversion efficiency as high as 3% has been achieved. A known
limitation like poor perovskite coverage over the mp-TiO2 surface,
revealed by ESEM analysis of the devices, could be a possible
reason for the relatively low performance of the solar cell. Taking
this fact into account, we believe that further optimizations in the
device fabrication together with a finer tuning of the CQDs
properties could promote a brand new type of HTM with very high
potential for perovskite solar cells.
Acknowledgments
We would like to thank ICIQ, Generalitat Catalunya, and Spanish
MINECO through Severo Ochoa Excellence Accreditation 2014–
2018 (SEV-2013-0319) and project CTQ2013-47183R for the
financial support. E.M-F. thanks the Spanish MINECO for the
Ramon y Cajal fellowship RYC-2010-06787 and funding through
the project MAT2012-31570. Eurecat financial aid is kindly
acknowledged.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.
synthmet.2016.04.025.
References
[1] X. Li, M. Rui, J. Song, Z. Shen, H. Zeng, Adv. Funct. Mater. 25 (2015) 4929–4947.
[2] L. Cao, X. Wang, M.J. Meziani, F. Lu, H. Wang, P.G. Luo, Y. Lin, B.A. Harruff, L.M.
Veca, D. Murray, S.-Y. Xie, Y.-P. Sun, J. Am. Chem. Soc. 129 (2007) 11318–11319.
[3] Y. Wang, A. Hu, J. Mater. Chem. C 2 (2014) 6921–6939.
[4] Y.-P. Sun, B. Zhou, Y. Lin, W. Wang, K.A.S. Fernando, P. Pathak, M.J. Meziani, B.A.
Harruff, X. Wang, H. Wang, P.G. Luo, H. Yang, M.E. Kose, B. Chen, L.M. Veca, S.-Y.
Xie, J. Am. Chem. Soc. 128 (2006) 7756–7757.
[5] R.C.E.R.a.a. http://www.nrel.gov/, (in).
[6] N. Ahn, D.-Y. Son, I.-H. Jang, S.M. Kang, M. Choi, N.-G. Park, J. Am. Chem. Soc. 137
(2015) 8696–8699.
[7] W.S. Yang, J.H. Noh, N.J. Jeon, Y.C. Kim, S. Ryu, J. Seo, S.I. Seok, Science 348
(2015) 1234–1237.
[8] N.J. Jeon, J.H. Noh, Y.C. Kim, W.S. Yang, S. Ryu, S.I. Seok, Nat. Mater. 13 (2014)
897–903.
[9] G.E. Eperon, V.M. Burlakov, P. Docampo, A. Goriely, H.J. Snaith, Adv. Funct.
Mater. 24 (2014) 151–157.
[10] J.M. Ball, M.M. Lee, A. Hey, H.J. Snaith, Energy Environ. Sci. 6 (2013) 1739–1743.
[11] L. Cabau, I. Garcia-Benito, A. Molina-Ontoria, N.F. Montcada, N. Martin, A.
Vidal-Ferran, E. Palomares, Chem. Commun. 51 (2015) 13980–13982.
22 S. Paulo et al. / Synthetic Metals 222 (2016) 17–22
[12] P. Qin, S. Paek, M.I. Dar, N. Pellet, J. Ko, M. Grätzel, M.K. Nazeeruddin, J. Am.
Chem. Soc. 136 (2014) 8516–8519.
[13] S. Kazim, F.J. Ramos, P. Gao, M.K. Nazeeruddin, M. Gratzel, S. Ahmad, Energy
Environ. Sci. 8 (2015) 1816–1823.
[14] K. Do, H. Choi, K. Lim, H. Jo, J.W. Cho, M.K. Nazeeruddin, J. Ko, Chem. Commun.
50 (2014) 10971–10974.
[15] Y. Zhi, L. Zhaohui, X. Minghan, M. Yujie, Z. Jing, S. Yanjie, G. Feng, W. Hao, Z.
Liying, Nano-Micro Lett. 5 (2013) 247–259.
[16] J.H. Im, C.R. Lee, J.W. Lee, S.W. Park, N.G. Park, Nanoscale 3 (2011) 4088–4093.
[17] D. Qu, M. Zheng, L. Zhang, H. Zhao, Z. Xie, X. Jing, R.E. Haddad, H. Fan, Z. Sun, Sci.
Rep. 4 (2014) 5294.
[18] X.T. Zheng, A. Ananthanarayanan, K.Q. Luo, P. Chen, Small 11 (2015)
1620–1636.
[19] B.J. Clark, T. Frost, M.A. Russell, UV Spectroscopy: Techniques, Instrumentation,
Data Handling, Chapman & Hall, London, New York, 1993.
[20] C. Reichardt, Solvent effects on the absorption spectra of organic compounds,
Solvents and Solvent Effects in Organic Chemistry, Wiley-VCH Verlag GmbH &
Co. KGaA, 2004, pp. 329–388.
[21] C. Wang, Z. Xu, C. Zhang, ChemNanoMat 1 (2015) 122–127.
[22] B. Omogo, J.F. Aldana, C.D. Heyes, J. Phys. Chem. C 117 (2013) 2317–2327.