PCCP

Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

View Article Online

PAPER View Journal | View Issue

Interpreting time-resolved photoluminescence

of perovskite materials†
Cite this: Phys. Chem. Chem. Phys.,
2020, 22, 28345
Emmanuel V. Péan,a Stoichko Dimitrov,b Catherine S. De Castroc and
Matthew L. Davies *ad

Time-resolved photoluminescence (TRPL) spectroscopy is a powerful technique to investigate excited

charge carrier recombinations in semiconductors and molecular systems. The analysis of the TRPL
decays of many molecular systems (e.g. molecules and organic materials) is usually fairly straightfoward
and can be fitted with an exponential function allowing extraction of the rate constants. Due to the
non-excitonic nature of charge carriers in lead halide perovskite materials coupled with the presence of
localised trap states in their band-gap, the TRPL of these materials is much more complicated to
interpret. Here we discuss two models used in the literature to simulate charge carrier recombinations
and TRPL in perovskites. These models consider the bimolecular nature of direct electron–hole recom-
bination but differ in their treatment of trap-mediated recombination with one model describing
trapping as a monomolecular process whereas the other as a bimolecular process between free carriers
Received 18th September 2020, and the available trap states. In comparison, the classical analysis of perovskite TRPL decay curves (using
Accepted 19th November 2020 a sum of exponentials) can lead to misinterpretation. Here we offer some recommendations for
DOI: 10.1039/d0cp04950f meaningful measurements of lead halide perovskite thin-films. The fluence dependence as well as
charge carrier accumulation due to incomplete depopulation of all photoexcited carriers between
rsc.li/pccp consecutive excitation pulses are discussed for both models.

Introduction concentration and the trapping rate constant is thus essential in

order to compare samples and devices, and improvement of
Since their first use in solar cells in 2009, lead halide perovskites perovskite solar cells.
have been employed in a wide variety of solar cell architectures Time-resolved photoluminescence (TRPL) spectroscopy is a
leading to an impressive 25.5% power conversion efficiency powerful technique that has been employed to study charge
(PCE) in 2020.1 Further improvement of the PCE may require carrier recombination kinetics of photoluminescent
enhancing the extraction of the excited charge carriers (electrons materials.8–10 TRPL measures the radiative electron–hole
and holes) from the perovskite layer to the electron and hole recombination after absorption of a short light excitation pulse
transport layers and reducing the recombination rate of these (typically of the order of the ps). The TRPL decay obtained is
carriers. Charge carrier recombination generally happens through affected by the different competitive radiative and non-
the direct recombination of an electron and a hole or mediated by radiative recombination processes happening in a system, as
defect-induced trap states.2 These trap states, through their for- well as its environment, and is therefore useful for extracting
mation and passivation, are also believed to be responsible for information related to these processes. TRPL is usually mea-
the instability and subsequent degradation of the perovskite.3–7 sured using time-correlated single photon counting (TCSPC) in
Extracting information related to these trap states such as their which the sample is periodically excited by a pulsed laser and
the sample emission is measured at the desired wavelength.
a
Ideally, after each excitation pulse, 0 or 1 photon is measured,
SPECIFIC IKC, Materials Research Centre, College of Engineering,
Swansea University Bay Campus, Fabian Way, Swansea SA1 8EN, UK.
thus generally requiring millions to billions of excitation pulses
E-mail: [email protected] to obtain an adequate signal to noise ratio (S/N). The TRPL
b
School of Biological and Chemical Sciences, Queen 58 Mary University of London, intensity ITRPL measured after P pulses is thus the sum of the
London E1 4NS, UK TRPL intensity IpTRPL after each excitation pulse p:
c
KAUST Solar Center KSC, Al-Kindi Bldg (5), Level 3, Room 3204-WS15,
Thuwal 23955-6900, Kingdom of Saudi Arabia
d
School of Chemistry and Physics, University of KwaZulu-Natal, Durban, X
P
p
South Africa ITRPL ðtÞ ¼ ITRPL ðt Þ (1)
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0cp04950f p¼1

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28345
 View Article Online

Paper PCCP

TRPL decays have been classically fitted with a sum of binding energy in perovskites, excited electrons and holes are
exponential decays Aiet/ti (Ai and ti are respectively the ampli- uncoupled and their recombination is bimolecular, that is a
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

tude and lifetime of component i):11 second order process with a rate dependent on both charge
X
n carriers concentrations.2,24,25 The contribution of this recombi-
ITRPL ðtÞ ¼ Ai et=ti (2) nation process to the change in the charge carrier concentra-
i tions over a small period of time dt is thus equal to the product
of the radiative band-to-band recombination rate constant kR,
When applied to perovskite materials, a bi-exponential
and the electron ne and hole nh concentrations:2,24,25
(n = 2) fit is generally done and the two components are usually    
assigned to trap-mediated recombination (shorter lifetime t1) dne dnh
¼ ¼ kR ne ðtÞnh ðtÞ (6)
and radiative recombination (longer lifetime t2).12–15 For per- dt R dt R
ovskite layers in contact with an electron or hole transport
material, the faster component has been linked to charge Photon recycling, during which emitted photons by the
carrier extraction from the perovskite to the transport system are re-absorbed before they escape it, has been
material.16–21 The amplitude of each exponential component previously observed in direct band-gap semiconductors with a
is usually attributed to the contribution of each process high absorption coefficient and a small Stokes shift (small
whereas the lifetime indicates the rate of the process.17,19 difference between emission and absorption maxima), such as
However, this model does not consider the dependence of the perovskites.2,26–28 As absorption is a much faster process than
recombination processes on the excited charge carrier concen- recombination (Bfs versus to Bns), carriers leading to photon
tration in perovskites. recycling can be assumed to not have recombined. The
Here, two models considering direct electron–hole recombi- observed radiative rate constant kR is thus a fraction of the
nation as well as trap-mediated recombination in homogenous internal radiative rate constant.2,26,29
perovskite systems are investigated. For most perovskite thin The TRPL intensity measured corresponds to the number of
films for solar cells (i.e. oca. 200 nm thick), carrier diffusion photons emitted by the system studied and is thus equal to:
can be assumed much faster than recombination such that the ITRPL(t) = kRne(t)nh(t) (7)
charge carriers can be considered homogenously distributed in
the film after photoexcitation (Note S1, ESI†). In such films, the It has been suggested that shallow trap state-mediated
photoexcited charge carrier concentration N0 generated by an recombination and trap-state mediated Auger recombination
excitation pulse of fluence I0 (in photons per cm2) is:22 may lead to non-radiative bimolecular recombination of charge
carriers at rate:2,25,26
I0 A    
N0 ¼ (3) dne dnh
D ¼ ¼ kN ne ðtÞnh ðtÞ (8)
dt N dt N
where A and D are respectively the film absorptance and
thickness. Since the carrier concentration N0 and the excitation where kN is the non-radiative bimolecular rate constant. As both
fluence I0 are proportional, they are used interchangeably in radiative (eqn (6)) and non-radiative (eqn (8)) bimolecular
this work. At room temperature, the excitons quickly dissociate recombination processes depend on the electron and hole
due to their low binding energy,23 leaving Dne electrons in the concentrations, they can be summed as the bimolecular
conduction band (CB) and Dnh holes in the valence band (VB). contribution of rate constant kB:
The total excited electron ne(t) and hole nh(t) concentrations in    
dne dnh
the conduction (CB) and valence (VB) bands at time t are then: ¼ ¼  ðkN þ kR Þ ne ðtÞnh ðtÞ (9)
dt B dt B |fflfflfflfflfflffl{zfflfflfflfflfflffl}
kB
ne(t) = Dne(t) + n0 (4)

nh(t) = Dnh(t) + p0 (5)

Trap-mediated recombination. Trap-mediated recombination,
where n0 and p0 are respectively the N-type and P-type doping also known as Shockley–Read–Hall recombination can be
concentrations which may originate from crystal impurities or described as a non-radiative bimolecular process between one
point defects, and (intrinsic carriers are ignored due to their carrier type and the available trap states.2,25 In the case of electron
low concentration – B104 cm3 at 300 K (Note S1, ESI†) – trapping, the trapping contribution is proportional to the product
compared to the doping and photoexcited concentrations of free electrons ne and available trap states concentrations NT–nt,
considered here). Here, we describe three of the recombination where NT is the concentration of trap states and nt is the
processes commonly considered in the literature: band-to-band, concentration of trapped electrons occupying these states:2,25
non-radiative bimolecular and trap-mediated recombination.    
dne dnt
¼ ¼ kT ne ðtÞ½NT  nt ðtÞ (10)
Recombination processes dt T dt T

Bimolecular recombination. Band-to-band recombination is where kT is the electron trapping rate constant. The detrapping of
a radiative process during which an excited electron directly an electron through its recombination with a hole in the VB can
recombines with a hole in the VB. Due to the low exciton be described as a non-radiative bimolecular process between

28346 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020
 View Article Online

PCCP Paper

trapped electrons and free holes:2,25 The BT model assumes the electron and hole concentration
    equal at all time (i.e. ne(t) = Dne(t) = nh(t) = Dnh(t) = n(t)) and
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

dnt dnh monomolecular trap-mediated recombination (Fig. 1a). The

¼ ¼ kD nt ðtÞnh ðtÞ (11)
dt D dt D rate equation for the BT model is thus build upon eqn (9)
and (12) which give:24,31,36
where kD is the detrapping rate constant. In the steady-state
regime and if the trap state concentration is much higher than dn
¼  kT nðtÞ  kB n2 ðtÞ (13)
the photoexcited carrier concentration, the electron and hole dt |fflfflffl{zfflfflffl} |fflfflfflffl{zfflfflfflffl}
T ðtÞ B ð tÞ
concentrations can be assumed equal at all time and trap-
mediated recombination can be described as a monomolecular where T(t) and B(t) are the trapping and bimolecular contribu-
processes:30–32 tions to the change in the charge carrier concentration at
    time t.
dne dnh
¼ ¼ kT ne ðtÞ ¼ kT nh ðtÞ (12) For TCSPC decay analysis, this differential equation is solved
dt T dt T
with the condition that each excitation pulse p generates N0
carriers which adds up to any carriers present in the bands just
Auger recombination is not discussed here as it has been
before the pulse. The initial carrier concentration np(t = 0) of
suggested that it only occurs at very high light fluences (above
pulse p is thus:
ca. 10 sun illumination) and should therefore be negligible in
PSCs under normal operation illumination conditions and in np ðt ¼ 0Þ ¼ N0 þ np1 ðRPÞ (14)
particular TRPL measurements conditions.33 Electron detrap-
ping back to the CB, which has been shown to lead to delayed where np1(RP) corresponds to the carrier concentration just
fluorescence (DF)34,35 could also influence the decay kinetics, before excitation pulse p and RP is the excitation repetition
however, our aim here is highlight complications with typical period. We assume that TCPSC experiments are started with a
analysis found in the literature where commonly DF is not film with no free carriers i.e. np=1(0) = N0 (this can be ensured by
considered. keeping the sample in the dark for long enough time so that
most carriers have recombined).
TRPL kinetic models As per eqn (7), the TRPL intensity measured as a function of
time after excitation pulse p is given by:
In perovskite films, both direct and trap mediated recombination
p
determine the overall charge recombination dynamics. Here we ITRPL ðtÞ ¼ kR ðnp ðtÞÞ2 (15)
discuss modelling of the TRPL decays with two possible models;
the bimolecular-trapping (BT) model and the bimolecular- Values of the different rate constants from the literature
trapping–detrapping (BTD) model (Fig. 1). Both models consider obtained using the BT model are reported in Table 1.
bimolecular recombination of charge carriers but differ in their Contrary to the BT model, the BTD model considers trapping
treatment of trap-mediated recombination. and detrapping as bimolecular processes (depicted in Fig. 1b) and

Fig. 1 Schematic of the recombination processes considered in the (a) bimolecular-trapping and (b) bimolecular-trapping–detrapping models. Excited
electrons (concentration ne) in the conduction band (CB) and holes (concentration nh) in the valence band (VB) undergo radiative and non-radiative
bimolecular recombination at rate constants kR and kN respectively (total rate constant kB). In (a), electrons and holes get trapped in trap states at a rate
constant kT. In (b), electrons get trapped in trap states (concentration NT) at a rate constant kT. Trapped electrons (concentration nt) can then detrap back
to the VB at a rate constant kD.

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28347
 View Article Online

Paper PCCP

Table 1 Recombination rate constants (bimolecular kB, trapping kT and detrapping kD), deep trap states (NT) and doping (p0) concentrations for different
perovskites obtained using the bimolecular-trapping (left) and bimolecular-trapping–detrapping (right) models [CH3NH3PbI3 (MAPI), CH3NH3PbI3(1x)Cl3x
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

(MAPIC), CH3NH3PbBr3 (MAPBr)], values taken from literature

Bimolecular-trapping model Bimolecular-trapping–detrapping model

Perovskite kB (1020 cm3 ns1) kT (103 ns1) Ref. kB (1020 cm3 ns1) kT (1020 cm3 ns1) NT (1012 cm3) p0 (1012 cm3) kD (1020 cm3 ns1) Ref.
MAPI 17 180 24 35 10 60 000 10 000 5 25
51–220 3–28 40
8.1 5 26 26–76 12 000 60 65 80 2*
92 14 41

MAPBr 12–20 2.9–1100 42 Not found

7.0 2.5 26

MAPIC 7.9 r0.5 26 49 B0 o500 1000 B0 25

as such it builds upon eqn (10) and (11). The electron and hole Values of the different rate constants obtained from the
concentrations are not always equal and so the rate equations literature using the BTD model are reported in Table 1.
for both carrier types and the trapped carrier concentration In this work, the use and limitations of the BT and BTD
needs to be solved. Within this model, trapping and detrapping models are discussed as well as lifetime interpretation of TRPL
is believed to be mediated by deep trap states while non- based on them.
radiative bimolecular recombination is mediated by shallow
trap states.2 In the case of electron trapping and considering
p-type doping (i.e. ne = Dne and nh = Dnh + p0) (as predicted for
Methodology
methylammonium lead iodide (MAPI) using first principle We describe here the methods and parameters used to simulate
calculations37), eqn (9)–(11) give the following set of rate the carrier concentrations and TRPL, as well as details about
equations:2,25,38,39 the fitting of the TRPL decays.
dDne Simulations
¼  kB Dne ðtÞ½Dnh ðtÞ þ p0   kT Dne ðtÞ½NT  nt ðtÞ (16)
dt |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
BðtÞ T ð tÞ Carrier concentrations and TRPL intensity were simulated by
following the process described in the block diagram in Fig. S1
dnt (ESI†), presented for the BT model only. The same process is
¼ kT Dne ðtÞ½NT  nt ðtÞ  kD nt ðtÞ½Dnh ðtÞ þ p0  (17) valid for the BTD model by replacing the equations with
dt |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
T ð tÞ DðtÞ eqn (16)–(20). Unless otherwise stated, all carrier concentra-
tions and TRPL decays were calculated after the first excitation
dDnh pulse only (referred to as the single pulse approximation in the
¼  kB Dne ðtÞ½Dnh ðtÞ þ p0   kD nt ðtÞ½Dnh ðtÞ þ p0  (18)
dt |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Results & discussions section). For the BT model, the carrier
BðtÞ DðtÞ
concentration was simulated using eqn (13) and (14) and kB =
where B(t), T(t) and D(t) are, respectively, the bimolecular, 50  1020 cm3 ns1 and kT = 15  103 ns1, values consistent
trapping and detrapping contributions to the change in the with what has been reported for perovskites (Table 1). The
carrier concentrations. TRPL intensity was then simulated using eqn (15). For the BTD
In TCSPC measurements, the initial carrier concentrations model, the carrier concentrations were simulated using
after excitation pulse p generating N0 carriers are: eqn (16)–(18) and the rate constants and the concentrations
reported for a pristine MAPI thin film in ref. 2 (* in Table 1):
Dnpe ðt ¼ 0Þ ¼ N0 þ Dnp1
e ðRPÞ kB = 26  1020 cm3 ns1, kT = 12 000  1020 cm3 ns1,
kD = 80  1020 cm3 ns1, p0 = 65  1012 cm3 and NT = 60 
Dnph ðt ¼ 0Þ ¼ N0 þ Dnp1
h ðRPÞ (19)
1012 cm3. The TRPL intensity was then simulated using eqn (20).
npt ðt ¼ 0Þ ¼ ntp1 ðRPÞ We note that there seems to be two scenarios in the literature, one
with fast trapping rate constant and smaller number of traps and
where Dnp1e (RP), Dnp1
h (RP) and nt
p1
(RP) are the electron, hole one with slower trapping rate constant and much larger number of
and trapped electron concentrations present in the system just traps. In this work we focus on the former as we feel it describes
before excitation pulse p. As for the BT model, it is assumed best the systems we have encountered. As discussed in the Results
that there are no carriers present in the system before the first & discussions section, TRPL decays resulting from billions
pulse i.e. Dnp=1 p=1 p=1
e (0) = Dnh (0) = N0, nt (0) = 0. of excitation pulses, as the ones experimentally measured, are
As per eqn (7), the TRPL intensity measured after an excita- simulated with only a few hundred excitation pulses until stabilisa-
tion pulse p is given by: tion of the carrier concentration(s) defined as:
p  
ITRPL ðtÞ ¼ kR Dnpe ðtÞ Dnph ðtÞ þ p0 (20) p 6
X (t)  nX(t)| r 10 N0
|np1 (21)

28348 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020
 View Article Online

PCCP Paper

Excitation pulse p is then assumed representative of the intensity after the first pulse within the single pulse approxi-
carrier concentration in the system after billions of pulses. This mation and after the last pulse determined from eqn 21
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

method is referred to as the multiple pulse approximation and within the multiple pulse approximation), as TRPL is usually
allows to greatly reduce the calculation cost. Further details experimentally measured until it reaches ‘‘zero’’. For the BT
about the calculations can be found in Note S2 (ESI†). model and the multiple pulse approximation, the repetition
The contribution C(t) of a recombination process (e.g. T(t) period was chosen such that ITRPL(RP)/ITRPL(0) = 0.25 in order to
for trapping) corresponds to the rate of charge carriers recom- lead to carrier accumulation. Prior to fitting, noise simulated
bined through such process at a given time t. Over a certain from a Poisson distribution was added to emulate experimental
% i.e. the total number of
period of time, the total contribution C, conditions (an amount of noise reasonably representative of
carriers recombined through this process, is given by: experimental conditions as shown in Fig. S4, ESI†) and the
ð decays were normalised with respect to their initial intensity.
C ¼ CðtÞdt (22) Fitting was carried using a least-square minimisation of the
free parameters (kB and kT for the BT model and kB, kT, kD, p0
The relative contribution %C(t) of a certain process corre- and NT for the BTD model) (Note S4, ESI†). We considered a
sponds to the ratio of carriers recombined through this process ‘‘best-case scenario’’ where the same values of the parameters
P
to the sum C ðtÞ of all carriers recombined at time t (Note S3, used in the simulated TRPL decays were used as guess values
ESI†): for the fitting optimisation.
C ðt Þ
%CðtÞ ¼ P (23)
C ðtÞ
Results & discussions
Consequently, we note %C% the relative total contribution
Here we discuss the application of the BT and BDT models to
which corresponds to the percentage of charge carriers recom-
describe TRPL decays of perovskite samples. First discuss the
bined through a certain process over a certain period of time.
excitation fluence dependence on TRPL after the first excitation
Since the carrier concentration is not the quantity observed
pulse (the superscript notation is ignored here for clarity i.e.
when measuring TRPL, it is necessary to quantify the contribu-
n p=1
X (t) = nX(t)). This is an important topic as TRPL decays are
tion of each recombination process to the changes in the TRPL
usually reported at a variety of fluences in the literature,
intensity. In the BT model (eqn (15)), the change in the TRPL
making comparisons across the literature difficult. We also
intensity over a small period of time dt is:
discuss the accuracy of the values retrieved from fitting these
dITRPL decays simulated with added noise at different excitation
¼  2kR nðtÞBðtÞ  2kR nðtÞT ðtÞ (24)
dt |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} fluences. In the second section, we discuss the importance of
BPL ðtÞ TPL ðtÞ
allowing recombination of all photoexcited charge carriers for
where BPL(t) and TPL(t) are the bimolecular and trapping fitting with a single excitation pulse and advise on how to avoid
contributions to the change in the TRPL intensity. In the BTD charge accumulation in the systems measured.
model (eqn (20)), the change in the TRPL intensity over a small
period of time dt is: Bimolecular-trapping model

dITRPL Fluence regimes. We investigate the fluence dependence of

¼  kR BðtÞ½Dne ðtÞ þ Dnh ðtÞ þ p0  the BT model by simulating the carrier concentration and TRPL
dt |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
BPL ðtÞ for a wide range of initial carrier concentrations N0. Within this
(25)
model, the carrier concentration normalised to its initial value
 kR T ðtÞ½Dnh ðtÞ þ p0   kR DðtÞDne ðtÞ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} n(t)/N0 (and therefore the normalised TRPL intensity) is
TPL ðtÞ DPL ðtÞ
observed to decay faster for increasing N0 (Fig. 2a and Note
where BPL(t), TPL(t) DPL(t) are respectively the bimolecular, S6, ESI†). In order to estimate the effect of trapping and
trapping and detrapping contributions to the change in the bimolecular recombination on the TRPL decay, we investigate
TRPL intensity. their contribution to the change in the TRPL intensity, %TPL
and %BPL respectively. In particular, we note that %TPL and
Fitting %BPL are respectively lower and higher than the contributions
In order to investigate the limitations of the models, fitting was to the change in the carrier concentrations %T% and %B% as the
carried out on TRPL decays simulated with the BT and BTD square of the carrier concentration favours bimolecular
models. With the BT model, the decays were simulated with 512 contribution, which means that bimolecular recombination
points (or channels) while 4096 points were used with the BTD has a greater effect on the TRPL profile than trapping (Fig. 2c).
model in order to resolve the fast trapping predicted by this Under low excitation fluence (here N0 o 1015 cm3), trap-
model. For both models, the excitation repetition period used ping is the dominant recombination process as kTN0 c kBN02.
(i.e. the time window used) was chosen such that the TRPL This is demonstrated in Fig. 2b where the trapping contribu-
intensity reaches ITRPL(RP)/ITRPL(0) = 104 with both single tion %TPL is close to 100% over the whole time-range studied
and multiple pulse approximations (where ITRPL is the TRPL (blue curves). Consequently, the total trapping contribution

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28349
 View Article Online

Paper PCCP
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

Fig. 2 Fluence dependence on the recombination process and the determination of their rate constants as predicted by the bimolecular-trapping
model. (a) Simulated TRPL decays, (b) relative trapping %TPL and bimolecular %BPL contributions to the change in the TRPL intensity and (c) total
bimolecular and trapping contributions to the change in the carrier concentration (%B% and %T) % and the TRPL intensity (%BPL and %TPL ) at increasing
excitation fluences (1014 cm3 r N0 r 1020 cm3). In (b) the red dashed lines indicate where the associated normalised TRPL intensity becomes lower
than 0.01 and 0.001. The red area defines the time for each excitation fluence over which the TRPL intensity is lower than 0.01 and thus where
information about trapping may not be recoverable. (d) Bimolecular kB and trapping kT rate constants obtained from fitting TRPL decays with added
noise, and normalised to the values used in the simulations (R2 4 0.998). A value of 1 or close to 1 indicates that it is accurately retrieved. The double
arrow indicates the fluence range where both rate constants can be retrieved with less than 15% relative error. The total trapping %TPL and bimolecular
%BPL contributions calculated from the fits are also shown as dashed lines.

%TPL is close to 100% while the total bimolecular contribution in Fig. 2b). The TRPL profile is dependent on the initial carrier
%BPL is negligible (Fig. 2c). In this regime, the BT rate equation concentration (green curves in Fig. 2a). Since both %TPL and
and TRPL (eqn (13) and (15)) can be simplified to: %BPL are non-negligible, it is possible to accurately retrieve
both recombination rate constants from the fitting of simu-
dn
 kT nðtÞ ) ITRPL ðtÞ / e2kT t (26) lated medium fluence TRPL decays with added noise as shown
dt
in Fig. 2d.
The TRPL therefore follows a mono-exponential trend Under high excitation fluence (kTN0 { kbN02, here N0 4
dependent only on the trapping rate constant kT and does not 10 cm3), the bimolecular contribution is initially dominant
19

depend on the initial carrier concentration N0 (blue curves in but quickly decreases as the carrier concentration n(t) decreases
Fig. 2a). The low contribution of bimolecular recombination in while the trapping contribution increases (yellow curves in
this regime may lead to inaccurate determination of the bimo- Fig. 2b). It may thus be possible to retrieve the trapping rate
lecular recombination rate constant kB when fitting TRPL. constant at the tail of the TRPL decay where the carrier
To illustrate this, we simulate TRPL decays at low fluence using concentration is low and %TPL is high. However, experimentally
the BT model and add some noise to emulate experimental measured TRPL decays usually include a certain degree of noise
conditions. The decays are then fitted using the same model which may limit the determination of the trapping rate con-
yielding the accurate value of kT but an inaccurate value of kB, stant as the S/N is lower at lower TRPL intensity (i.e. at the tail
even though the correct values of the parameters were used as of the decay). For example, the time at which the normalised
initial guesses (Fig. 2d). TRPL intensity reach 0.01 is shown as dashed red lines in
Under medium excitation fluence (kTN0 B kBN02, here N0 B Fig. 2b. Under high fluence, trapping only contributes a few
10 –1019 cm3), the bimolecular and trapping contributions
15
percent of the total recombinations before the TRPL reaches
are comparable, with bimolecular recombinations dominant in 0.01 normalised intensity in Fig. 2b (yellow curves). Consequently,
the initial decay and trapping dominant in its tail (green curves fitting TRPL decays simulated at high excitation fluences with

28350 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020
 View Article Online

PCCP Paper

added noise yields an accurate value of kB but not of kT (Fig. 2d). When measured with a 200 ns repetition period, np=1(RP) =
Improving the S/N (e.g. 0.001 curve in Fig. 2b) improves the 0.01N0 carriers are left in the bands just before the second
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

accuracy of the trapping rate determined (Fig. S5, ESI† shows excitation pulse (red curve). The initial carrier concentration of
the value of kT and kB retrieved from fitting for different amount of the second excitation pulse is thus np=2(0) = np=1(RP) + N0 =
noise). 1.01N0 leading to a similar variation of the carrier concen-
The accuracy of the rate constant values retrieved from tration as after the first pulse (Fig. 3a). The resulting TRPL
fitting (at any fluence) can be estimated by calculating the total decay after 109 pulses is thus similar to the TRPL decay
bimolecular and trapping contributions using these values. For simulated with only the first excitation pulse (Fig. 3b) and
example, in Fig. 2d the values of the rate constant retrieved single pulse fitting can therefore be used. Conversely, with a
from fitting become inaccurate when %BPL and %TPL get close 25 ns repetition period, np=1(RP) = 0.34N0 carriers are left in the
to zero, respectively at low and high fluence, consistent with system before the second excitation pulse (purple curve in
our previous observations. Fig. 3a). The higher initial carrier concentration of the second
The different fluence regimes of the BT model are sum- pulse np=2(0) = 1.34N0 accelerates the recombination due to the
marised in Table S1 (ESI†). higher bimolecular contribution. The TRPL decay obtained
after 109 pulses is 1.9 times more intense than the single pulse
TRPL decay (not shown here) and decays faster due to the
Incomplete recombination higher carrier concentration (Fig. 3b). Such decay cannot be
As TCSPC measurements generally require millions to billions fitted with a single pulse and would require the simulation of
excitation pulses to obtain an appropriate S/N, fitting experi- all the excitation pulses. The accumulation of charge carriers in
mentally measured TRPL by simulating only the first excitation the system continues until as many carriers recombine after an
pulse (referred to as the single pulse approximation hereafter) excitation pulse as carriers are generated by the latter (Fig. 3a):
requires that all charge carriers recombine between two consecu-
tive excitation pulses. To illustrate this, we compare the evolution np(0) = np(RP) = N0 (27)
of the carrier concentration after multiple consecutive pulses with
excitation repetition periods RP = 25 ns and RP = 200 ns (Fig. 3a). This carrier accumulation (CA) does not affect the profile of the
In both cases, the first pulse excites np=1(0) = N0 carriers. TRPL decay at low fluences (as shown in Fig. 3c & Fig. S6, ESI†)

Fig. 3 Effect of incomplete depopulation on the carrier concentration and TRPL over time as predicted by the bimolecular-trapping model at medium
fluence excitation (N0 = 1017 cm3) and different excitation repetition periods (RP = 200 ns, 25 ns). (a) Carrier concentration np(t) normalised to its initial
value after the first excitation pulse np=1(0) over 400 ns and (b) normalised TRPL intensity after 109 pulses. The TRPL decay simulated after a single pulse is
also shown for comparison. (c) Maximum difference between the single pulse and multiple pulse normalised TRPL and carrier concentration at increasing
N0 (the repetition period was chosen such that ITRPL(RP)/ITRPL(0) = 0.25).

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28351
 View Article Online

Paper PCCP

as monomolecular recombination does not depend on the Since the TRPL intensity is proportional to the square of the
carrier concentration (although the increased initial carrier carrier concentration, eqn (28) implies that the carrier concen-
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

concentration may drive the system in a regime where trations also reach zero:
bimolecular recombination are non-negligible). As the fluence
is increased, CA becomes more and more important partially IpTRPL(RP) = 0 ) np(RP) = 0 (29)
due to the higher contribution of bimolecular recombination
As per eqn (27), this means that the initial carrier concen-
(Fig. 3c).
tration of each pulse is np(0) = N0 and therefore CA is negligible.
Assuming that the carrier concentration stabilises over a
It is thus straightforward to avoid CA according to the BT model
small fraction of the total number of excitation pulses experi-
by choosing an excitation repetition period such that the TRPL
mentally required to obtain a good S/N TRPL decay (as sug-
decay is fully measured (e.g. when ITRPL is at least 0.01% it
gested by our simulations), it is possible to simulate and fit
indicates under 1% of non-recombined charge carriers).
TRPL decays resulting from billions of excitation pulses by only
simulating a small fraction of them and by considering the last
Bimolecular-trapping–detrapping model
simulated pulse to be representative of the system measured
(referred to as the multiple pulse approximation hereafter). Contrary to the BT model, the BTD model does not consider
However, multiple pulse fitting is much slower than single trapping and detrapping to be synchronous. For the following
pulse fitting and measuring incomplete decays (i.e. without analysis and discussion, we assume that the doping concen-
their tail) reduces the range of fluences where kT and kB can tration is of the same order of magnitude than the trap state
both be accurately retrieved as discussed in Note S7 (ESI†). concentration, and that the trapping rate constant is much
It is thus essential to avoid CA in order to use single pulse higher than the bimolecular and detrapping rate constants.
fitting. This can be ensured by choosing an excitation repetition
period such that the TRPL decay reaches zero. The TRPL decay Fluence regimes
can only reach zero if the TRPL intensity after each pulse I pTRPL We investigate the fluence dependence of the BTD model by simu-
reaches zero by the end of the repetition period as per eqn (1): lating the carrier concentration and TRPL for a wide range of initial
carrier concentrations. Contrary to the BT model, the BTD model
ITRPL(RP) = 0 - IpTRPL(RP) = 0 (28)
does not predict a straightforward trend of the TRPL profile (Fig. 4a).

Fig. 4 Fluence effect on the recombination processes predicted by the bimolecular-trapping–detrapping model. (a) TRPL decays and (c) total
bimolecular, trapping and detrapping contributions to the change in the carrier concentrations (%B, % %T,% %D
% respectively, full curves) and TRPL intensity
(%BPL , %TPL , %DPL respectively, dashed curves) at increasing excitation fluences (1012–1017cm3). (b) Evolution of the photoexcited electron (Dne), hole
(Dnh) and trapped electron (nt) concentrations, and TRPL intensity (ITRPL), and (d) associated relative bimolecular (%B), trapping (%T) and detrapping (%D)
contributions after an excitation pulse generating N0 = 2  1014 cm3 (NT: trap state concentration).

28352 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020
 View Article Online

PCCP Paper

We note that over the whole fluence range studied, the total initial intensity and its intensity at tsat where the TRPL decrease
trapping %T% and detrapping %D% contributions are always equal slows down, multiplied by the initial carrier concentration
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

since as many carriers get trapped and detrapped, however, their (Note S10, ESI†):
contribution to the change in the TRPL intensity are different as
discussed hereafter (Fig. 4c). NT B [ITRPL(0)  ITRPL(tsat)]N0 (33)
Under low excitation fluence (i.e. N0 { NT, here N0 o 5 
1012 cm3), the hole concentration can be assumed constant The trapping rate constant can also be extracted from the
and equal to the doping concentration, and therefore the trapping phase by fitting the initial TRPL decay with a mono-
bimolecular contribution has a monomolecular behaviour: exponential function, as long as it remains mono-exponential
(i.e. that the trap states remain mostly empty, Note S10, ESI†).
B(t) = kBDne(t)[Dnh(t) + p0] E kBDne(t)p0 (30) As the trap states are filled, the detrapping and trapping
contributions become equal such that any electron detrapped
As the photoexcited electron concentration is much lower gets replaced by another electron and the trap states remain full
than the trap states concentration, the trap states can be at any time (Fig. 4b and d). This steady-state continues until the
considered empty at all time and the trapping contribution electron concentration becomes low enough that trapping
has also a monomolecular behaviour: becomes slower than detrapping and the trap states depopu-
T(t) = kTDne(t)[NT  nt(t)] E kTDne(t)NT (31) late. As for the low fluence regime, detrapping can happen after
the TRPL intensity has reached zero and has therefore a limited
The bimolecular and trapping processes are therefore undis- contribution to the change in its intensity as shown in Fig. 4c.
tinguishable in this regime. Since the hole concentration is Since all the recombination processes are non-negligible in this
constant, the TRPL intensity follows the same trend as the regime, it is possible to accurately retrieve all the parameters of
electron concentration and is mono-exponential which shape the model (Note S9, ESI†). However, contrary to the BT model
does not depend on the excitation fluence (Fig. 4a): where a set of values gives a very distinct TRPL decay, the BTD
model can yield very similar TRPL decays calculated from
ITRPL(t) = p0Dne p e(kBp0+kTNT)t = ekMt (32) different sets of values, even over a wide range of fluences as
where kM = kBp0 + kTNT. In the present simulations, we assumed discussed in Notes S11 & S12 (ESI†). It is thus necessary to
kTNT c kBp0 and therefore the TRPL decays only depend on the ensure that only 1 set of values can fit the data by running the
product kTNT (other cases are discussed in Note S8, ESI† but fitting optimisation with a wide range of guess values such
eqn (32) remains valid). This is consistent with the negligible that all the optimisations converge toward the same solution
bimolecular %BPL and detrapping %DPL contributions to the (Note S11, ESI†).
change in the TRPL intensity shown in Fig. 4c. As discussed for Under high excitation fluences (i.e. N0 c NT, here N0 4 6 
the BT model, only kM can be retrieved from fitting TRPL decays 1015 cm3), the trap states are immediately saturated and
in this regime and therefore any combination of values of kB, most carriers undergo bimolecular recombination leading to
p0, kT and NT yielding the same value kM are a solution to the fit. dominant bimolecular contribution (Fig. 4c). If the doping
We demonstrate this by fitting TRPL decays simulated with the concentration is much lower than the photoexcited carrier
BTD model in the low fluence regime where different sets of concentrations, only the bimolecular rate constant can be
values of kB, p0, kT and NT are retrieved, all yielding the same kM accurately retrieved (Notes S9 & S11, ESI†).
value (Note S9, ESI†). Therefore, we advise to measure TRPL decays in the medium
As the excitation fluence increases and become close to the fluence regime where all the recombination processes are non-
trap state concentration, the availability of the trap states needs negligible. This regime can easily be identified by the fast
to be considered and trapping has a bimolecular behaviour initial drop of the TRPL intensity followed by a slower decay.
leading to non-exponential TRPL profiles. TRPL decays measured in the low and high fluence regime can
Under medium excitation fluence (i.e. NT o N0 o B100NT, respectively be used to determine the product kTNT and kB.
here N0 B 6  1013–6  1015 cm3), the electrons quickly fill the Furthermore, we recommend carrying out several fittings with
trap states to saturation (Fig. 4b). The trapping then slows different sets of guess values in order to confirm that only 1 set
down due to the absence of available trap states and the free of values can fit the TRPL decay (Note S11, ESI†).
electrons can only undergo bimolecular recombination while
the high trapped electron concentration leads to an increase of Incomplete recombination
the detrapping contribution (Fig. 4d). As these processes are As previously discussed, experimental TRPL can only be fitted
slower than trapping, the TRPL decay decrease slows down after with a single excitation pulse by ensuring recombination of all
about 100 ns in Fig. 4b. Since the hole concentration can be the carriers between consecutive excitation pulses. As for the BT
assumed constant over the initial trapping phase (as previously model, CA due to incomplete electron recombination can easily
discussed in the low fluence regime) the TRPL intensity can be avoided by ensuring that the TRPL intensity reaches zero,
thus be considered proportional to the electron concentration however, the BTD model predicts that trapped electrons can
as observed in Fig. 4b. The trap state concentration can thus recombine non-radiatively after the free electron concentration, and
be roughly determined from the difference between the TRPL therefore after the TPRL intensity, has reached zero. For example, in

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28353
 View Article Online

Paper PCCP

Fig. 5a and b although both RP = 100 ms and RP = 15 ms allow complete depopulation of all trap states by gradually increasing
recombination of most free electrons, RP = 15 ms does not allow the repetition period. Doing so allows more and more trapped
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

all the trapped electrons and free holes to recombine before the electrons to detrap and the TRPL profile converges toward the
next excitation pulse. The resulting TRPL after many consecu- single pulse TRPL profile (Fig. 5c).
tive excitation pulses shows a smaller initial decrease due to the As for the BT model, the multiple pulse approximation can
presence of the trapped electrons, effectively reducing the be used to fit TRPL decays subjected to CA. However, the higher
available trap state concentration in the model (Fig. 5b and c). occupation of the trap states limits the trapping contribution
Furthermore, the higher initial hole concentration affects and thus may reduce the accuracy of the trapping rate constant
bimolecular recombination. and the trap state concentration retrieved (Fig. S9c, ESI†).
Due to the non-radiative nature of trap-mediated Our observations on the BTD model are summarised in
recombination, CA of trapped electrons is more complicated Table S2 (ESI†).
to avoid than CA due to incomplete depopulation of the free
electrons. The maximum difference between the single and Determination of the doping concentration
multiple pulse TRPL decays was calculated at different
Although the absolute intensity measured by TCSPC is not
excitation fluences for a repetition period chosen such that
usable due to the unknown ratio between the number of
ITRPL(RP)/ITRPL(0) = 104 (Fig. 5d). Trap based CA does not affect
photons measured and photons emitted, it can still potentially
the TRPL profile in the low fluence regime as the latter does not be used to determine the doping concentration of the system.
depend on the initial carrier concentration (however, if the The experimentally measured initial TRPL intensity depends
initial carrier concentration becomes comparable to the trap
upon the initial carrier concentration (assuming negligible CA)
state concentration, the TRPL profile then becomes dependent
and the doping concentration:
on the carrier concentration as shown in Fig. 5d). In the high
fluence regime, trap based CA is always negligible as trapping ITRPL(0) p N0(N0 + p0) (34)
is also negligible compared to bimolecular recombination
(Fig. 5d). The highest difference due to CA is observed in the The doping concentration can be determined by measuring
medium fluence regime. In this regime, it is possible to ensure the initial TRPL intensity at two or more fluences NA0 and NB0

Fig. 5 Evolution of the photoexcited electron Dne, hole Dnh, trapped electron nt concentrations and TRPL intensity ITRPL as predicted by the
bimolecular-trapping–detrapping model (N0 = 1014 cm3) over 3 consecutive excitation pulses for a (a) RP = 100 ms repetition period leading to
negligible carrier accumulation and a (b) RP = 15 ms repetition period leading to non-negligible carrier accumulation (X = Dne, Dnh, nt, ITRPL).
(c) Normalised TRPL decays after multiple excitation pulses for different excitation repetition periods, compared to the TRPL after the first excitation pulse
(single pulse). (d) Maximum difference between the single pulse and multiple pulse normalised TRPL and carrier concentration decays at increasing N0
(the repetition period was chosen such that ITRPL(RP)/ITRPL(0) B 104).

28354 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020
 View Article Online

PCCP Paper
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

Fig. 6 Lifetime-based interpretation of TRPL decays. (a) TRPL decays simulated with the bimolecular-trapping model with increasing trapping rate
constant kT and (b) corresponding total trapping %T% and bimolecular %B% contributions to the change in the carrier concentration. The amplitudes Ai
and lifetimes ti obtained from bi-exponential fitting of the TRPL decays are shown in (b) and (c) respectively (R2 4 0.999).

(Note S13, ESI†): The resulting amplitudes and lifetimes from bi-exponential fitting
    of these decays are reported in Fig. 6b and c. The amplitude A1
B A A 2 A B B 2
ITRPL ð0ÞP N0 ITRPL ð0ÞP N0 of the shorter exponential decreases while the amplitude A2 of the
p0 ¼ A
(35)
ITRPL ð0ÞPB N0B  ITRPL
B ð0ÞPA N0A longer component increases as the trapping rate constant
increases. Considering the general assignment of the shorter
where the number of pulses P can be calculated from the
component to trap-mediated recombination and the longer com-
measurement time M and the repetition period:
ponent with bimolecular recombination, the results of this fit
M could be interpretated as an increase of the contribution
P¼ (36)
RP of bimolecular recombination and a decrease of the trapping
contribution. However, this is inconsistent with our simulations
as an increase of the trapping contribution and a decrease of the
Lifetime-based interpretation of TRPL bimolecular contribution are predicted (Fig. 6b). Although the
Interpretation of TRPL decays based on their lifetime is highly decrease of the shorter lifetime t1 is consistent with the higher
popular thanks to its simplicity. However, this method presents trapping rate, the decrease of t2 is inconsistent with the slower
many limitations and drawbacks. Lifetime-based analysis does bimolecular rate due to lower carrier concentration induced by
not consider the dependence of the recombination processes the higher trapping rate (Fig. 6c). Bi-exponential fitting of TRPL
on the carrier concentration unlike the BT and BTD models. It decays can thus lead to misinterpretation of the processes occur-
can therefore only be used to compare samples measured with ring in the system studied and should therefore be used with
the same initial carrier concentration N0, not subjected to any scarcity.
CA (as it affects the initial concentration) and presenting
the same recombination processes (e.g. it is not possible to Conclusions
compare a perovskite film with and without contact with an
electron transport layer as injection may happen in the latter). Charge carrier recombination in perovskites is complex and as
Furthermore, any observation made using this method is only a result several models for charge dynamics have been used in
valid for the initial carrier concentration used and may not be the literature to extract meaningful data from TRPL decays. In
valid under solar cell operation for example. If these requirements this work, the limitations and fluence dependence of 2 charge
are met and assuming that the electron and hole concentrations carrier recombination models commonly used in the literature
are equal at all time (as in the BT model), a longer overall lifetime were reviewed. These models consider radiative and non-
of the TRPL indicates a longer lifetime of the free carriers and radiative bimolecular recombination, partially responsible
thus potentially a better efficiency of the perovskite when working for the dependence of the TRPL profile on the carrier concen-
in a complete device. trations, but differ in their treatment of trap-mediated
Multi-exponential models are usually used to quantify recombination.
recombination processes by associating lifetimes with them. Both models predict dominant monomolecular trapping at
Here we discuss the suitability of this type of analysis by fitting low excitation fluence leading to a mono-exponential TRPL
TRPL decays simulated with the BT model with a bi-exponential decay. Since other recombination processes are negligible in
function. Based on TRPL decays fitted with a bi-exponential this regime, only trapping related information can be retrieved
function typical in literature, TRPL decays were simulated with (that is the trapping rate constant kT for the BT model and
the BT model with an increasing trapping rate constant (Fig. 6a). the product of the trapping rate constant and the trap state

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28355
 View Article Online

Paper PCCP

concentration kTNT for the BTD model). Under high fluence, increase the complexity of the models and may require more
bimolecular recombination is dominant and only the complex measurements.
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

bimolecular rate constant kB can be accurately retrieved. Due to the dependence of the TRPL on excitation fluence and
Finally, in the medium fluence regime, the BT model predicts excitation repetition period, we recommend these parameters be
non-negligible trapping and bimolecular recombination included in the experimental details of all published work.
allowing the accurate measurement of the rate constants of We hope that this work will aid in the analysis of the TRPL of
both processes. In the same regime, the BTD model predicts a perovskite materials and will help the investigation of their
fast initial trapping phase leading to a fast decrease of the TRPL complex and interesting properties.
intensity, which can be used to roughly determine the trap state
concentration and trapping rate constant, followed by a slower
decrease. The BTD model can be simplified by determining the Conflicts of interest
values of some parameters beforehand; we propose a method There are no conflict to declare.
to calculate the doping concentration from the initial TRPL
intensity measured in the medium fluence regime. We demon-
strate that the validity of the values retrieved from the BT model Acknowledgements
can be ensured by calculating the contribution of the processes
to the change in the TRPL intensity. Due to the higher complexity We are grateful for the financial support of the EPSRC
of the BTD model, we advise ensuring that multiple sets of (EP/R016666/1 and EP/S001336/1) and both the EPSRC and
parameters’ guess values for the fitting optimisation yield the Innovate UK for the SPECIFIC Innovation and Knowledge
same solution. We note that we limited this study on fitting single Centre and the European Regional Development Fund through
TRPL decays showing that it may be possible to accurately retrieve the Welsh Government for support to the Sêr Solar program.
results from the model used in the medium fluence regime. MLD is grateful for funding through the EPSRC GCRF SUNRISE
However, we expect the accuracy of the results would be further project (EP/P032591/1). We would like to acknowledge the
improved by carrying out a global fitting of TRPL decays measured assistance provided by the European Regional Development
at different excitation fluences. Fund through the Welsh Government (80708), the Ser Solar
Incomplete depopulation of all photoexcited charge carriers project via Welsh Government and Edinburgh Instruments.
within the excitation repetition period leads to an accumula-
tion of charge carriers in the system. This carrier accumulation References
usually happens over less than a few hundred excitation pulses
and can alter the intensity and profile of the TRPL decay. 1 National Renewable Energy Laboratory. Best Research-Cell
Carrier accumulation due to incomplete electron recombina- Efficiency Chart|Photovoltaic Research|NREL [Internet],
tion was found to be easily avoidable by choosing an excitation National Renewable Energy Laboratory, 2019 [cited 2020
repetition period such that the TRPL intensity reaches zero. Sep 18]. Available from: https://www.nrel.gov/pv/cell-
However, carrier accumulation due to incomplete detrapping efficiency.html.
as predicted by the BTD model only is harder to avoid as it can 2 R. Brenes, D. Guo, A. Osherov, N. K. Noel, C. Eames and
happen after the TRPL intensity has reached zero. In the E. M. Hutter, et al., Metal Halide Perovskite Polycrystalline
medium fluence regime, complete detrapping can be ensured Films Exhibiting Properties of Single Crystals, Joule, 2017,
by comparing the TRPL decay profile measured with 2 excita- 1(1), 155–167. Available from: http://www.cell.com/joule/
tion repetition periods. For both models, we showed that fitting pdf/S2542-4351(17)30031-4.pdf.
TRPL decay subjected to carrier accumulation by simulating 3 S. G. Motti, D. Meggiolaro, A. J. Barker, E. Mosconi,
multiple pulses is possible but may be more limited than fitting C. A. R. Perini and J. M. Ball, et al., Controlling competing
TRPL decays presenting negligible carrier accumulation using photochemical reactions stabilizes perovskite solar cells,
single pulse fitting. Nat. Photonics, 2019, 13(8), 532–539. Available from: http://
Finally, lifetime interpretation of TRPL of perovskites can www.nature.com/articles/s41566-019-0435-1.
only be used in specific cases and quantifying recombination 4 J. S. W. Godding, A. J. Ramadan, Y.-H. Lin, K. Schutt,
processes using a multi-exponential fit may yield incorrect H. J. Snaith and B. Wenger, Oxidative Passivation of Metal
interpretations of the TRPL. Halide Perovskites, Joule, 2019, 3(11), 2716–2731. Available
Both models discussed here have advantages and dis- from: https://www.sciencedirect.com/science/article/abs/pii/
advantages: the BT model is a more simplistic approach of S2542435119303812.
recombinations in perovskites but allows easy extraction of the 5 Z. Andaji-Garmaroudi, M. Anaya, A. J. Pearson and
process rate constants while the BTD model is more complex S. D. Stranks, Photobrightening in Lead Halide Perovskites:
but requires more complex measurements in order to accu- Observations, Mechanisms, and Future Potential, Adv. Energy
rately retrieve the parameter values and avoid trap based carrier Mater., 2020, 10(13), 1903109. Available from: https://onlineli
accumulation. Further processes can be considered in these brary.wiley.com/doi/abs/10.1002/aenm.201903109.
models e.g. electron detrapping back the CB, Auger recombina- 6 N. Aristidou, C. Eames, I. Sanchez-Molina, X. Bu, J. Kosco
tion and carrier diffusion. However, these processes will likely and M. Saiful Islam, et al., Fast oxygen diffusion and iodide

28356 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020
 View Article Online

PCCP Paper

defects mediate oxygen-induced degradation of perovskite 18 Z. Li, J. Dong, C. Liu, J. Guo, L. Shen and W. Guo, Surface
solar cells, Nat. Commun., 2017, 8, 15218. Available from: Passivation of Perovskite Solar Cells Toward Improved
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

http://www.nature.com/doifinder/10.1038/ncomms15218. Efficiency and Stability, Nano-Micro Lett., 2019, 11(1), 50.

7 A. Leguy, Y. Hu, M. Campoy-Quiles, M. I. Alonso, O. J. Weber Available from: http://link.springer.com/10.1007/s40820-
and P. Azarhoosh, et al., Reversible hydration of 019-0282-0.
CH3NH3PbI3 in films, single crystals and solar cells, Chem. 19 K.-C. Hsiao, M.-H. Jao, B.-T. Li, T.-H. Lin, S. H.-C. Liao and
Mater., 2015, 27, 3397–3407. Available from: http://pubs.acs. M.-C. Wu, et al., Enhancing Efficiency and Stability of Hot
org/doi/abs/10.1021/acs.chemmater.5b00660. Casting p–i–n Perovskite Solar Cell via Dipolar Ion Passivation,
8 U. Noomnarm and R. M. Clegg, Fluorescence lifetimes: ACS Appl. Energy Mater., 2019, 2(7), 4821–4832. Available from:
fundamentals and interpretations, Photosynth. Res., 2009, https://pubs.acs.org/doi/10.1021/acsaem.9b00486.
101(2–3), 181–194. Available from: http://www.ncbi.nlm.nih. 20 J. Tao, N. Ali, K. Chen, Z. Huai, Y. Sun and G. Fu, et al.,
gov/pubmed/19568954. Enhanced efficiency in perovskite solar cells by eliminating
9 T. Kirchartz, J. A. Márquez, M. Stolterfoht and T. Unold, the electron contact barrier between the metal electrode
Photoluminescence-Based Characterization of Halide and electron transport layer, J. Mater. Chem. A, 2019
Perovskites for Photovoltaics, Adv. Energy Mater., 2020, , 7(3), 1349–1355. Available from: http://xlink.rsc.org/?DOI=
1904134. Available from: https://onlinelibrary.wiley.com/ C8TA10630D.
doi/abs/10.1002/aenm.201904134. 21 Z. Yao, W. Wang, H. Shen, Y. Zhang, Q. Luo and X. Yin,
10 J. R. Lakowicz, Principles of Fluorescence Spectroscopy, ed. et al., CH3NH3PbI3 grain growth and interfacial properties
J. R. Lakowicz, Springer US, Boston, MA, 2006. Available in meso-structured perovskite solar cells fabricated by two-
from: http://link.springer.com/10.1007/978-0-387-46312-4. step deposition, Sci. Technol. Adv. Mater., 2017, 18(1),
11 J. S. S. De Melo, T. Costa, C. S. De Castro and A. L. Maçanita, 253–262. Available from: http://www.ncbi.nlm.nih.gov/
Photophysics of fluorescently labeled oligomers and polymers, pubmed/28458747.
Photochemistry, 2013, 41, 59–126. 22 Y. Yang, D. P. Ostrowski, R. M. France, K. Zhu, J. van de
12 Z.-Y. Y. Zhang, H.-Y. Y. Wang, Y.-X. X. Y. Zhang, Y.-W. Lagemaat and J. M. Luther, et al., Observation of a hot-
W. Hao, C. Sun and Y.-X. X. Y. Zhang, et al., The role of phonon bottleneck in lead-iodide perovskites, Nat. Photonics,
trap-assisted recombination in luminescent properties of 2016, 10(1), 53–59. Available from: http://www.nature.com/
organometal halide CH3NH3PbBr3 perovskite films and articles/nphoton.2015.213.
quantum dots, Sci. Rep., 2016, 6(1), 27286. Available from: 23 L. M. Herz, Charge-Carrier Dynamics in Organic-Inorganic
http://www.nature.com/articles/srep27286. Metal Halide Perovskites, Annu. Rev. Phys. Chem., 2016,
13 T. Du, J. Kim, J. Ngiam, S. Xu, P. R. F. Barnes and 67(1), 65–89. Available from: http://www.annualreviews.
J. R. Durrant, et al., Elucidating the Origins of Subgap Tail org/doi/10.1146/annurev-physchem-040215-112222.
States and Open-Circuit Voltage in Methylammonium Lead 24 Y. Yamada, T. Nakamura, M. Endo, A. Wakamiya and
Triiodide Perovskite Solar Cells, Adv. Funct. Mater., 2018, Y. Kanemitsu, Photocarrier recombination dynamics in
28(32), 1801808. Available from: http://doi.wiley.com/10. perovskite CH3NH3PbI3 for solar cell applications, J. Am.
1002/adfm.201801808. Chem. Soc., 2014, 136(33), 11610–11613. Available from:
14 J. Yang, Y. Yu, L. Zeng, Y. Li, Y. Pang and F. Huang, et al., http://pubs.acs.org/doi/10.1021/ja506624n.
Effects of aromatic ammoniums on methyl ammonium lead 25 E. M. Hutter, G. E. Eperon, S. D. Stranks and T. J. Savenije,
iodide hybrid perovskite materials, J. Nanomater., 2017, Charge Carriers in Planar and Meso-Structured Organic-
2017, 1–6. Available from: https://www.hindawi.com/jour Inorganic Perovskites: Mobilities, Lifetimes, and Concen-
nals/jnm/2017/1640965/. trations of Trap States, J. Phys. Chem. Lett., 2015, 6(15),
15 P. Caprioglio, F. Zu, C. M. Wolff, J. A. Márquez Prieto, 3082–3090. Available from: http://pubs.acs.org/doi/10.1021/
M. Stolterfoht and P. Becker, et al., High open circuit acs.jpclett.5b01361.
voltages in pin-type perovskite solar cells through strontium 26 J. M. Richter, M. Abdi-Jalebi, A. Sadhanala, M. Tabachnyk,
addition, Sustainable Energy Fuels, 2019, 3(2), 550–563. J. P. H. Rivett and L. M. Pazos-Outón, et al., Enhancing
Available from: http://xlink.rsc.org/?DOI=C8SE00509E. photoluminescence yields in lead halide perovskites by
16 J. Yin, J. Cao, X. He, S. Yuan, S. Sun and J. Li, et al., photon recycling and light out-coupling, Nat. Commun.,
Improved stability of perovskite solar cells in ambient air 2016, 7(1), 13941. Available from: http://www.nature.com/
by controlling the mesoporous layer, J. Mater. Chem. A, 2015, articles/ncomms13941.
3(32), 16860–16866. Available from: http://xlink.rsc.org/ 27 D. W. deQuilettes, K. Frohna, D. Emin, T. Kirchartz,
?DOI=C5TA02843D. V. Bulovic and D. S. Ginger, et al., Charge-Carrier Recombi-
17 P.-W. Liang, C.-Y. Liao, C.-C. Chueh, F. Zuo, S. T. Williams nation in Halide Perovskites, Chem. Rev., 2019, 119(20),
and X.-K. Xin, et al., Additive Enhanced Crystallization 11007–11019.
of Solution-Processed Perovskite for Highly Efficient 28 T. W. Crothers, R. L. Milot, J. B. Patel, E. S. Parrott, J. Schlipf
Planar-Heterojunction Solar Cells, Adv. Mater., 2014, 26(22), and P. Müller-Buschbaum, et al., Photon Reabsorption
3748–3754. Available from: http://doi.wiley.com/10.1002/ Masks Intrinsic Bimolecular Charge-Carrier Recombina-
adma.201400231. tion in CH3NH3PbI3 Perovskite, Nano Lett., 2017, 17(9),

This journal is © the Owner Societies 2020 Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 | 28357
 View Article Online

Paper PCCP

5782–5789. Available from: https://pubs.acs.org/doi/10. 36 Y. Yamada, H. Yasuda, T. Tayagaki and Y. Kanemitsu,

1021/acs.nanolett.7b02834. Temperature dependence of photoluminescence spectra of
Published on 23 November 2020. Downloaded by INDIAN INSTITUTE OF TECHNOLOGY BOMBAY on 12/25/2023 5:53:25 PM.

29 L. M. Pazos-Outón, M. Szumilo, R. Lamboll, J. M. Richter, nondoped and electron-doped SrTiO3: Crossover from auger
M. Crespo-Quesada and M. Abdi-Jalebi, et al., Photon recycling recombination to single-carrier trapping, Phys. Rev. Lett.,
in lead iodide perovskite solar cells, Science, 2016, 351(6280), 2009, 102(24), 247401. Available from: http://www.ncbi.nlm.
1430–1433. Available from: http://science.sciencemag.org/con nih.gov/pubmed/19659046.
tent/351/6280/1430. 37 D. Meggiolaro, S. G. Motti, E. Mosconi, A. J. Barker and
30 Y. Yamada, H. Yasuda, T. Tayagaki and Y. Kanemitsu, J. Ball, Andrea Riccardo Perini C, et al. Iodine chemistry
Photocarrier recombination dynamics in highly excited determines the defect tolerance of lead-halide perovskites,
SrTiO3 studied by transient absorption and photolumines- Energy Environ. Sci., 2018, 11(3), 702–713. Available from:
cence spectroscopy, Appl. Phys. Lett., 2009, 95(12), 121112. http://xlink.rsc.org/?DOI=C8EE00124C.
Available from: http://aip.scitation.org/doi/10.1063/1.3238269. 38 S. D. Stranks, V. M. Burlakov, T. Leijtens, J. M. Ball,
31 T. Handa, D. M. Tex, A. Shimazaki, A. Wakamiya and A. Goriely and H. J. Snaith, Recombination Kinetics in
Y. Kanemitsu, Charge Injection Mechanism at Hetero- Organic-Inorganic Perovskites: Excitons, Free Charge, and
interfaces in CH3NH3PbI3 Perovskite Solar Cells Revealed Subgap States, Phys. Rev. Appl., 2014, 2(3), 034007. Available
by Simultaneous Time-Resolved Photoluminescence and from: https://link.aps.org/doi/10.1103/PhysRevApplied.2.
Photocurrent Measurements, J. Phys. Chem. Lett., 2017, 034007.
8(5), 954–960. Available from: http://pubs.acs.org/doi/10. 39 T. J. Savenije, D. Guo, V. M. Caselli and E. M. Hutter,
1021/acs.jpclett.6b02847. Quantifying Charge-Carrier Mobilities and Recombination
32 A. R. S. Kandada, V. D’Innocenzo, G. Lanzani and A. Petrozza, Rates in Metal Halide Perovskites from Time-Resolved
Photophysics of Hybrid Perovskites. in Unconventional Thin Microwave Photoconductivity Measurements, Adv. Energy
Film Photovoltaics, Cambridge, Royal Society of Chemistry, Mater., 2020, 10, 1903788. Available from: http://arxiv.org/
2016, ch. 4, p. 107–140. Available from: http://ebook.rsc.org/ abs/2001.02569.
?DOI=10.1039/9781782624066-00107. 40 Y. Li, W. Yan, Y. Li, S. Wang, W. Wang and Z. Bian, et al.,
33 A. Alnuaimi, I. Almansouri and A. Nayfeh, Performance of Direct Observation of Long Electron–Hole Diffusion Dis-
planar heterojunction perovskite solar cells under light tance in CH3NH3PbI3 Perovskite Thin Film, Sci. Rep., 2015,
concentration, AIP Adv., 2016, 6(11), 115012. Available from: 5(1), 14485. Available from: http://www.nature.com/articles/
http://aip.scitation.org/doi/10.1063/1.4967429. srep14485.
34 V. S. Chirvony, S. González-Carrero, I. Suárez, R. E. Galian, 41 C. Wehrenfennig, G. E. Eperon, M. B. Johnston, H. J. Snaith
M. Sessolo and H. J. Bolink, et al., Delayed Luminescence in and L. M. Herz, High charge carrier mobilities and lifetimes
Lead Halide Perovskite Nanocrystals, J. Phys. Chem. C, 2017, in organolead trihalide perovskites, Adv. Mater., 2014,
121(24), 13381–13390. Available from: https://doi.org/10. 26(10), 1584–1589. Available from: http://doi.wiley.com/10.
1021/acs.jpcc.7b03771. 1002/adma.201305172.
35 V. S. Chirvony and J. P. Martı́nez-Pastor, Trap-Limited 42 B. Wu, H. T. Nguyen, Z. Ku, G. Han, D. Giovanni and
Dynamics of Excited Carriers and Interpretation of the N. Mathews, et al., Discerning the Surface and Bulk Recom-
Photoluminescence Decay Kinetics in Metal Halide Perovs- bination Kinetics of Organic-Inorganic Halide Perovskite
kites, J. Phys. Chem. Lett., 2018, 9(17), 4955–4962. Available Single Crystals, Adv. Energy Mater., 2016, 6(14), 1600551.
from: https://doi.org/10.1021/acs.jpclett.8b01241. Available from: http://doi.wiley.com/10.1002/aenm.201600551.

28358 | Phys. Chem. Chem. Phys., 2020, 22, 28345--28358 This journal is © the Owner Societies 2020