Can Tauc plot extrapolation be used for direct-band-gap semiconductor nanocrystals?

Y. Feng, S. Lin, S. Huang, S. Shrestha, and G. Conibeer

Citation: Journal of Applied Physics 117, 125701 (2015); doi: 10.1063/1.4916090

View online: http://dx.doi.org/10.1063/1.4916090
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/117/12?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Semiconducting graphene nanoribbon retains band gap on amorphous or crystalline SiO2
Appl. Phys. Lett. 99, 183103 (2011); 10.1063/1.3657494

Size dependent optical band gap of ternary I-III-VI 2 semiconductor nanocrystals

J. Appl. Phys. 105, 073106 (2009); 10.1063/1.3103768

Band gap energy modifications observed in trivalent In substituted nanocrystalline Sn O 2

Appl. Phys. Lett. 90, 233117 (2007); 10.1063/1.2746407

Size-dependent energy band gap and dielectric constant within the generalized Penn model applied to a
semiconductor nanocrystallite
J. Appl. Phys. 100, 084301 (2006); 10.1063/1.2357421

Band gap in epitaxial NaCl-structure CrN(001) layers

J. Appl. Phys. 91, 5882 (2002); 10.1063/1.1466528

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 JOURNAL OF APPLIED PHYSICS 117, 125701 (2015)

Can Tauc plot extrapolation be used for direct-band-gap semiconductor

nanocrystals?
Y. Feng,a) S. Lin, S. Huang, S. Shrestha, and G. Conibeer
School of Photovoltaic and Renewable Energy Engineering, UNSW Australia, Sydney, NSW 2052, Australia
(Received 21 January 2015; accepted 9 March 2015; published online 23 March 2015)
Despite that Tauc plot extrapolation has been widely adopted for extracting bandgap energies of
semiconductors, there is a lack of theoretical support for applying it to nanocrystals. In this paper,
direct-allowed optical transitions in semiconductor nanocrystals have been formulated based on a
purely theoretical approach. This result reveals a size-dependant transition of the power factor used
in Tauc plot, increasing from one half used in the 3D bulk case to one in the 0D case. This size-
dependant intermediate value of power factor allows a better extrapolation of measured absorption
data. Being a material characterization technique, the generalized Tauc extrapolation gives a more
reasonable and accurate acquisition of the intrinsic bandgap, while the unjustified purpose of
extrapolating any elevated bandgap caused by quantum confinement is shown to be incorrect.
C 2015 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4916090]
V

I. INTRODUCTION continuous by accounting for the smearing of discrete levels.

The second part explicitly deals with the rate of direct optical
The absorptance measurement has played an important
transitions for electrons excited from the valence band to the
role in characterizing the optical band gap for bulk and thin
conduction band (interband). The theoretical derivation
film semiconductors.1–3 The theoretical foundation of this
results in size-dependent formulae for the absorption-energy
technique (Tauc plot extrapolation4) is based on the energy
dependence of the above-band-gap absorption, which dependence, followed by extrapolation of a few sets of ex-
appears either as a squared relation (direct-allowed-transition perimental data using the generalized Tauc approach.
dominated) or a square-rooted relation (indirect-allowed-
transition dominated).5 However, for semiconductor nano-
crystals, there is still a lack of clear and solid theoretical rela- II. ELECTRONIC DENSITY OF STATES IN
tions between the absorption coefficient and the photon NANOCRYSTALS
energy. In fact, without justifying its feasibility for low- For direct-band-gap materials that have parabolic band
dimensional structures, researchers have adopted the stand- shapes at the edges of the conduction and valence bands, car-
ard Tauc extrapolation technique widely for the purpose of riers are normally treated as free particles with “effective”
retrieving the optical band gaps of nanocrystals.6–9 masses. The nanocrystal is assumed to be spherically sym-
Practically, the linearity of the Tauc plot used in nanocrystals metric, with its radius denoted by R. The electron affinity is
is usually not as good as bulk materials, and therefore the
denoted by V. When treating free electrons, V is the finite
extrapolation can be arbitrary. This urges a solid theoretical
barrier height at the interface of the nanocrystal. This leads
support to either justify this technique or generalize the form
to electron wave functions leaked into the atmosphere. The
of Tauc extrapolation, so that it can be safely applied to
time-independent Schr€odinger equation under the effective
semiconductor nanocrystals.
mass approximation (EMA) is
For nanocrystals, one should expect the absorption spec-
trum deviates from that of bulk crystals, partly because of  
h2
the discretization and blueshift of the energy spectrum of   D þ Ec ðrÞ /ðrÞ ¼ E/ðrÞ: (1)
2m
electronic states. But what is more relevant to this paper is
the relaxation of the momentum selection rule, which no lon- Equation (1) proposes a eigenstate problem for a single
ger holds in nanocrystals with sufficiently small size. These electron in a spherical symmetric potential field, with the
two factors together can cause a deviation from the squared mass of free electron replaced by the effective mass of car-
power law for direct-allow transitions, and a size-dependent riers, m . While the angular parts of the eigenfunctions are
transition is expected, which creates a “grey regime” trivially spherical harmonics, Ylm ðh; uÞ, the radial parts are
between a fully periodic crystalline structure and an expressed in terms of spherical Bessel function of the 1st kind
infinitesimal-sized dot. Accordingly, the theoretical part of ð1Þ
jl ðxÞ and spherical Hankel function of the 1st kind hl ðxÞ10
this paper is organized into two parts. The first part discusses 
the size-dependent density of states in semiconductor nano- Ajl ðkrÞ; r < R
RðrÞ ¼ ð1Þ (2)
crystals, treating the electronic state spectrum as quasi- Bhl ðijrÞ; r > R;
a)
Electronic mail: [email protected] where

0021-8979/2015/117(12)/125701/7/$30.00 117, 125701-1 C 2015 AIP Publishing LLC

V

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 125701-2 Feng et al. J. Appl. Phys. 117, 125701 (2015)

rffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2m E 2m ðV0  EÞ states. In fact, most absorption measurements are performed
k¼ ; j¼ : (3) for a collection of large amount of nanocrystals, instead of a
h2 h2
single one. For realistic nanocrystal dispersions, discrete
The coefficients, A and B, are determined by the conti- energy levels have small spacings and usually get smeared
nuity of the wave function and the normalization condition. due to a small degree of size non-uniformity. Assuming a
Here, we only discuss the confined states, (E < V0 ). The con- quasi-continuous spectrum, the accumulated number of
dition of logarithmic derivative gives quantized eigen- states nlm ðvÞ (for specified quantum numbers l and m) can be
energies. For practically fabricated nanocrystals, the radius R solved from the continuous version of (7)
and the barrier height V0 are sufficiently large to have
pffiffiffiffiffiffiffiffiffiffiffiffiffi  
ð1Þ dnlm 1 1 1
2m V0 R   h, so that jl ðkRÞ and hl ðijRÞ may be replaced ¼ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 1 : (10)
by their asymptotic forms. In fact, the barrier height V0 is dv Dv p u2  v2
normally of the order of 10 eV, and the radius R is of the
pffiffiffiffiffiffiffiffiffiffiffiffiffi
order of 10 nm; this provides 2m V0 R  102 h. The asymp- Substituting the dimensionless integration variable, dv, by
totic forms of the spherical Bessel and Hankel functions are the energy interval, d, we integrate (10) over energy
expressed by trigonometric functions,
nlm ðEÞ  nlm ðE0 Þ ¼ PðEÞ  PðE0 Þ; (11)
 
1 lþ1
jl ðkRÞ  cos kR  p ; where E is the current energy level and E0 is the ground-
kR 2 state energy (for a specified quantum number l). The func-
   (4)
ð 1Þ 1 lþ1 tion P(E) is defined by the integration
hl ðijRÞ  exp  jR  p :
jR 2 ð E rffiffiffiffiffiffi  
m R d 1
PðEÞ¢ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ 1
The continuity condition of the logarithmic derivative, 0 2 h p u2  v2
8 9
  < " pffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi# sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2 =
ð Þ 1 V0  V0  E 8m R E
1 djl ðkrÞ  1 hl 1 ðijr Þ  ¼ 4 tan1 pffiffiffi þ :
¼ ; (5) 2p : h2 ;
jl ðkRÞ dr r¼R hð1Þ ðijRÞ dr r¼R E
(12)
is written in terms of the asymptotic forms
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Since the ground-state energy E0 corresponds to the first so-
kRcotðkRÞ ¼ u2  ðkRÞ2 ; (6) lution v0 of (6), we can safely assume v0
u, which after
pffiffiffiffiffiffiffiffiffiffiffiffiffi some algebraic manipulation approximates Eq. (12) to
where the dimensionless parameter u¢ 2m V0 R=h. n hpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i o
1
Equation (6) is used to solve the short-wavelength limits of PðE0 Þ ¼ 4 tan1 cot2 v0 þ 1  cotv0 þ 2v0
eigenenergies for a specified angular quantum number l. 2p  
Before discussing on the density of the corresponding eigen- 1 1
 4 ðp  v0 Þ þ 2v0 ¼ 1 : (13)
states in a finite-size spherical particle, let us define dimen- 2p 2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sionless parameters v¢kR and f ðvÞ¢ ðu=vÞ2  1. Substituting (13) into (11) and noticing that the accumulated
Supposing vn and vnþ1 are two successive solutions of (6), af- number of states at the first node (E0) is always equal to 1,
ter some algebraic manipulation we have we obtain n0m ðEÞ  PðEÞ whose validity relies on the as-
ymptotic form.4 As we are only interested in energy levels
f ðvnþ1 Þ  f ðvn Þ ¼ ½cotðvnþ1 Þ  cotðvn þ pÞ: (7) well below the barrier height (i.e., deeply bounded states,
where E
V0 ),pthe ffiffiffiffiffiffiffiffifirst
ffi term of (12)
pffiffiffiis negligible. In this
The first-order approximation of (7) is case, PðEÞ  2k m E, where k¢R= 2ph. Accordingly, the
density of states for a specified set of quantum numbers
f 0 ½ðvnþ1 þ vn Þ=2ðvnþ1  vn Þ ¼ csc2 ½ðvnþ1 þ vn Þ=2 (l and m) is
 ðvnþ1  vn  pÞ: (8)  rffiffiffiffiffiffi
dPðEÞ  m
q0 ðÞ ¼   k : (14)
Substituting (6) into (8) gives rise to the expression for the dE E¼ 
interval of neighboring states in the space spanned by the
dimensionless parameter v Equation (14) illustrates the state distribution for a
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi specified set of quantum numbers in the asymptotic regime

2 ( is large enough to have kR  1). To obtain a total state
p u2  ðvnþ1 þ vn Þ=2
Dvn ¢vnþ1  vn ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2 : (9) distribution, stationary states with all possible angular
u2  ðvnþ1 þ vn Þ=2 þ 1 momenta need to be counted. According to the asymptotic
form shown in (4), jl ðkRÞ is a periodic function with respect
In this paper, we assume a quasi-continuous absorption to l, if kR  1, i.e., jl ðkRÞ ¼ jl4 ðkRÞ. This implies that the
spectrum in the energy range of consideration due to a vari- difference in their accumulated numbers of states
ety of smearing mechanisms. This assumption allows us to [Pl ðEÞ  Pl4 ðEÞ] is a constant in the asymptotic regime.
derive a continuous function for the density of electronic The property of the Bessel functions tells us that this

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 125701-3 Feng et al. J. Appl. Phys. 117, 125701 (2015)

constant is equal to 2.11 This means that (5) has less nodes in integral, h/k0 l0 m0 j/klm i, can be evaluated by substituting elec-
the regime, where the asymptotic form (4) is not yet valid tronic bound states, j/klm i ¼ RðrÞYlm ðh; uÞ, where the radial
(kR  1), if reducing l to l – 4. part R(r) is given by Eq. (2)
For states with a zero angular momentum (l ¼ 0), the ðR
accumulated number of states, n0m ðEÞ, is completely deter- h/k0 l0 m0 j/klm i ¼ AA0 r2 jl0 ðk0 rÞjl ðkrÞdr
mined by (14), and it is possible to integrate it from the 0
ground state energy, E0, of the nanocrystal. Noted that ð p ð 2p
n0m ðE0 Þ ¼ 1. After determining the number of states with  Yl0 m0 ðh; uÞYlm ðh; uÞ sin hdudh;
0 0
l ¼ 0, we can obtain the number of states contributed by posi-
(20)
tive angular quantum numbers using the relation
l l where A and A0 are the respective normalization factors for
nlm ðEÞ ¼ n0m ðEÞ  ¼ PðEÞ  : (15) the state j/klm i and j/k0 l0 m0 i.
2 2
Due to the orthogonal property of the spherical harmon-
Taking into account, the energy degeneracy of the magnetic ics, the angle-dependent integral in (20) is equal to dl;l0 dm;m0
quantum number, m, the total accumulated number of states, (Kronecker d-function), while the integration in the radial
N(E), and the corresponding density of states, N 0 ðEÞ part is evaluated using the identity properties of Bessel
¼ dNðEÞ=dE, are functions
ð EÞ
ðR
X
2P
N ð EÞ ¼ ð2l þ 1Þnlm ðEÞ r 2 jl0 ðk0 rÞjl ðkrÞdr
0
l¼0
k0 R2

1 ¼ j ðkRÞ jl0 1 ðk0 RÞ  jl0 þ1 ðk0 RÞ

¼ PðEÞ½2PðEÞ þ 1½4PðEÞ þ 1 ; (16) 2ðk2  k02 Þ
l
6
kR2

N 0 ðEÞ ¼ QðEÞq0 ðEÞ; (17)  jl0 ðk0 RÞ jl1 ðkRÞ  jlþ1 ðkRÞ : (21)
ð 2
2 k k 02 Þ
where q0 ðEÞ is expressed by (14), and the function Q(E) is
Substituting the asymptotic forms of the Bessel functions4
defined by
into (21),
1 ðR
QðEÞ ¢ 4P2 ðEÞ þ 2PðEÞ þ : (18)
6 r 2 jl0 ðk0 rÞjl ðkrÞdr
0
 
1 sin½ ðk  k0 ÞR sin½ ðk þ k0 ÞR  lp
III. INTERBAND ABSORPTION IN NANOCRYSTALS  
2kk0 k  k0 k þ k0
To evaluate the absorption profile, the matrix elements sin½ ðk  k ÞR0
of the light perturbing Hamiltonian mq A p^ are to be calcu-  : (22)
2kk0 ðk  k0 Þ
lated first. Here, q and m denote the charge and mass of a
^ is the
free electron, A is the vector potential of light, and p It is noted that both approximations are valid, if kR  1 and
momentum operator for electron. In the framework of effec- k0 R  1. For a normalized wave function, the probability
tive mass approximation (or envelop function approximation within the sphere shall equal to unity. Substituting (22) into
equivalently), the actual electron wave function, U, within 2
R=2k2 ¼ 1. This leads to the
(20), we have h/klm j/klm i ¼pAffiffiffiffiffiffiffiffi
the nanocrystal can be decomposed into the product of a normalization factor A ¼ k 2=R. Finally, the general form
fast-varying Bloch wave, w, and a slow-varying envelop of the inner product h/k0 l0 m0 j/klm i is expressed by
function / [the eigenfunction solved in (1)]. This decompo-
sition allows evaluation of the perturbation matrix elements sin½ ðk  k0 ÞR
h/k0 l0 m0 j/klm i ¼ dl;l0 dm;m0 : (23)
using the following scheme: ðk  k 0 ÞR
D q  E h q
  Back to the absorption problem, the band-to-band transi-
Uk0 l0 m0  A p
^ Uklm ¼ A ðhwjwih/k0 l0 m0 jr/klm i
m im tion rate CðÞ associated with a specified photon energy,
þ hwjrwih/k0 l0 m0 j/klm iÞ; (19)  þ Eg , is expressed below, following the perturbation theory
ð ð X
where k is the wave-vector defined in (3), while l and m 2p  
C ð Þ ¼ dc dv qclm ðc Þqvl0m0 ðv Þ
denote the integers of spherical harmonics, or equivalently h 0 0 0
lm;l m 0
the angular quantum number and the magnetic quantum D q  E2
   
number following the terminology in atomic physics. The   Uk0 l0 m0  A p^ Uklm  dðc þ v  Þ; (24)
m
vector potential, A, has negligible variation over the dimen-
sion of the nanocrystal, due to the long wavelength of light where c and v are the excess photon energy, , distributed
(1 lm). Since /klm varies much slower than wn (i.e., to the conduction band and to the valence band, respectively.
jr/klm j
jrwj), the first term in (19) is negligible com- The energy conservation law gives rise to the Dirac d-func-
pared to the second term. In the second term, the overlap tion dðc þ v  Þ, which is part of the Fermi’s golden rule.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 125701-4 Feng et al. J. Appl. Phys. 117, 125701 (2015)

qclm ðc Þ and qvlm ðv Þ are the respective densities of states in Two extreme cases of (29) are considered here, corre-
the conduction band and in the valence band for a given set sponding to a 3D bulk semiconductor and a 0D (ultra small)
of l and m. They are given by (14) only within the asymptotic semiconductor dot.
regime (kR  1). Since k and k0 are the wave-vectors of Case I: A bulk semiconductor corresponds to the limit
conduction-band electrons and valence-band holes, respec- that the radius of the nanocrystal becomes infinite (R ! 1).
tively, they are replaced by more systematic symbols kc and At this limit, we have
kv. According to (3), they are defined by
rffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffi sin2 ½ ðkc  kv ÞR pdðkc  kv Þ
lim ¼
2me c 0 2mh v R!1 ðkc  kv Þ2 R2 R
kc ¢k ¼ 2
; kv ¢k ¼ ; (25) pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
h h2 ph 2me mh 
¼ dðc  c0 Þ: (31)
where me and mh are the respective effective masses of elec-
3
Rðme þ mh Þ2
trons and holes. Now substituting (19) and (23) into (24),
ð The d-function in (31) corresponds to the direct transition in
2phq2 0 2 bulk semiconductors (i.e., an electron being excited from the
C ð Þ ¼ jA hw jrwij dc Njoint ðc ;   c Þ
m2 0 valence band to the conduction band with its crystal momen-
sin2 ½ ðkc  kv ÞR tum being conserved). Substituting (31) into the general tran-
 ; (26) sition rate expression (29),
ðkc  kv Þ2 R2
pffiffiffi
where the joint number of states Njoint ðc ;   c Þ is defined 2pAhme mh pffiffi
lim CðÞ ¼ 3 : (32)
by R!1 Rðme þ mh Þ2
X
Njoint ðc ;   c Þ¢ qclm ðc Þqvlm ð  c Þ Case II: At the 0D limit, the nanocrystal radius R ! 0. The
lm trigonometric term in (29) has infinitely slow variation. Its limit
¼ minfQc ðc Þ; Qv ð  c Þgqc0 ðc Þqv0 ð  c Þ:
sin2 ½ ðkc  kv ÞR
(27) lim ¼ 1: (33)
R!0 ðkc  kv Þ2 R2
Defined by (14), the functions qc0 ðc Þ and qv0 ðv Þ share the
same form except for their respective effective masses. In a This result implies that unlike bulk materials, the transitions
similar way, Qc ðc Þ and Qv ðv Þ are both defined by (18), which occurring in 0D dots do not necessarily follow the condition
characterise the “energy distributions” of spherical harmonics of crystal momentum conservation. In fact, the effect of this
in the conduction band and in the valence band, respectively. selection rule is completely eliminated. The total transition
The monotonic function, Fðc Þ ¼ Qc ðc Þ  Qv ð  c Þ, has rate at the 0D limit is
only one zero crossing within the regime ð0; Þ. If we denote
lim CðÞ
this particular energy by c0 ; CðÞ, it can be evaluated sepa- R!0
2 0sffiffiffiffiffiffi1 0sffiffiffiffiffiffi1 3
rately via two integrations: ½0; c0  and ½c0 ; .
As we are interested only in energies, E, that are large 6 me A mh A pffiffiffiffiffiffiffiffiffiffiffi

7
¼ A4mh tan1 @  þ me tan1 @  me mh 5:
enough to fall into the asymptotic regime (kR  1), P(E) is mh me
also much larger than unity so only the highest-order term is
required in (18). Therefore, QðÞ  4P2 ðÞ  ð4kÞ2 m , (34)
which gives
According to Case I and Case II, at the two extreme cases
m  (R ! 1 and R ! 0), the total transition rate obeys the form of
c0 ¼  h  : (28) a power law, which forms the basis of Tauc plot extrapolation
me þ mh
ð þ Eg Það þ Eg Þ / CðÞ / n ; (35)
The total transition rate shown in (26) is then evaluated to be
( ð rffiffiffiffiffiffiffiffiffiffiffiffi 2 where the total energy of photon,  þ Eg , is decomposed into
c0
 c sin ½ ðkc  kv ÞR
C ð Þ ¼ A m e dc a bandgap energy, Eg, and an excess carrier kinetic energy, .
0    c ðk c  k v Þ2 R 2 aðEÞ is the absorption coefficient of nanocrystals at energy E
ð rffiffiffiffiffiffiffiffiffiffiffiffi )
  c sin2 ½ ðkc  kv ÞR and n denotes a constant power factor (Tauc factor). At the

þ mh dc ; (29) bulk limit (R ! 1), the Tauc factor n degenerates to its
c0 c ðkc  kv Þ2 R2 standard value 1/2, while at the 0D limit n is increased to 1.
Except for the two extreme cases, generally the total
where transition rate CðÞ expressed in (29) does not have the form
shown in (35). However, for energies at the vicinity of the
4 pffiffiffiffiffiffiffiffiffiffiffi 2pc
2
A ¼ ð2kÞ me mh ; band edge, (35) can be used to approximate the energy de-
h
 2 (30) pendence described by (29) and (1). The remaining task is to
hq find the most appropriate Tauc factor, n, which yield a suffi-
c2 ¼ jA hwjrwij2 :
m ciently small discrepancy between (35) and (29).

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 125701-5 Feng et al. J. Appl. Phys. 117, 125701 (2015)

By choosing these two representative material systems, it is

demonstrated that Tauc fitting scheme with corrected factor,
n, can be used in both solid-state and solution-based NCs.
SnO2 as a direct-gap material normally has an optical
bandgap of around 3.88 eV.12 From Fig. 2(c), we find that
the standard direct-gap tauc factor (1/2) cannot provide a sat-
isfied fitting result from both aspects: the Tauc plot is curved
downwards rather than being linear, and its intersection sig-
nificantly overestimates the band gap. On the other hand, the
zero-dimensional limit (R ! 0) of the Tauc factor, which is
equal to unity, underestimates the band gap [Fig. 2(d)]
though the linearity is no longer a problem. Compared with
the results fitted by the power factor of either one or one
half, the size-dependent intermediate power factors do dem-
FIG. 1. Direct-allowed optical transition rate vs phonon energy above the onstrate a better linearity and a more consistent and accurate
band gap for different sizes of nanocrystals (represented by different colors
extrapolation for the bandgap value [2(b)]. The power factors
and dot shapes). The dots show the exact values of transition rates which are
calculated by (29), while the curves are approximate transition rates used in Fig. 2(b) are optimised by fitting the power-law
obtained by fitting the exact rates to the power law equation (35). As shown approximation (35) to the exact absorption expression (29).
in this figure, the optimal power factor increases from one half to one when One should note that the Tauc extrapolation was performed
increasing the nanocrystal radius from 0.1 nm to 500 nm.
by deliberately neglecting the absorption spectrum close to
the absorption edge. This is because the extrapolation relies
Figure 1 demonstrates the relations between the transi-
on the assumption of quasi-continuous assumption, which is
tion rate, C, and the photon energy, , described by (29) for
no longer valid at low energies due to quantum confinement.
different particle sizes. It is shown that each curve fits well
Additionally, the disturbance from the amorphous tail and
with a Tauc plot with a suitable power factor n. This means exciting emission peak needs to be avoided.
that the power law (35) can approximate the exact absorption Figure 2(a) demonstrates a similar trend with Fig. 1. The
expression (29) with sufficient accuracy, if with a proper optimal power factor n for SnO2 decreases from 1 to 1/2
size-dependent power factor. with the size of QD. In addition to the least-square-error
(LSE) optimal power factor, the mean relative error of the
IV. EXPERIMENTAL ANALYSIS USING THE MODIFIED power-law approximation has been illustrated in Fig. 2(a),
TAUC METHOD which is defined as
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
+ffi
To show the practical implication intrigued by our u*
u Cexact  Capprox 2
theory, Tauc plots on experimental absorption data with dif- D ¼ t   ;
 (36)
ferent power factors, n, are presented in this section. Two Cexact
quantum-dot materials are discussed here, including SnO2
nanocrystals (NCs) embedded in solid-state SiO2 matrix and where Cexact is given by the exact relation (29) while Capprox
wet-chemical synthesized colloidal solution of Ag2S NCs. is given by the power-law approximation (35). It is noted that

FIG. 2. (a) LSE power factor for SnO2

obtained by fitting the power-law
approximation to the exact formula for
optical transition and the relative error
of such fitting. (b) Tauc plot extrapola-
tion for SnO2 NC with size-dependent
power factors. (c) Tauc plot extrapola-
tion for SnO2 NC with a constant
power factor of one half; (d) Tauc plot
extrapolation for SnO2 NC with a con-
stant power factor of one. The extrapo-
lation does not account for low-energy
part of absorption spectrum, in order to
avoid discrete energy levels, amor-
phous tail, and excitonic emission.

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 125701-6 Feng et al. J. Appl. Phys. 117, 125701 (2015)

FIG. 3. (a) LSE power factor for Ag2S

obtained by fitting the power-law
approximation to the exact formula for
optical transition and the relative error
of such fitting. (b) Tauc plot extrapola-
tion for Ag2S NC with size-dependent
power factors. (c) Tauc plot extrapola-
tion for Ag2S NC with a constant
power factor of one half. (d) Tauc plot
extrapolation for Ag2S NC with a con-
stant power factor of one. The extrapo-
lation does not account for low-energy
part of absorption spectrum, in order to
avoid discrete energy levels, amor-
phous tail, and excitonic emission. It is
noted that the sharp drop at 1.07 eV
is due to insensitive facility calibration
at the signal peak of the solvent (chlo-
roform). It does not affect the spectrum
region used for Tauc plot, which is
away from any solvent peaks.

the maximum error of adopting the power law occurs at the bulk material, as long as a power factor of one half is
QD diameter of around 0.4 nm [Fig. 2(a)]. At such a diameter adopted. However, this bandgap value is an incorrect overes-
the fitted power factor demonstrates a local minimum, which timation and therefore cannot be used to determine the blue-
is believed to be due to the inaccuracy of the power law, i.e., shifted bandgap value quantitatively. In fact, both Figs. 2
10% deviation from the exact formula (29). Generally, the and 3 demonstrate incorrect orders of bandgap values, for
transition of the power factor should show a monotonous trend one should expect a monotonous decrease of bandgap value
from unity to one half with the NC diameter. with the size of QD.
As a direct-gap chalcogenide material, Ag2S NC From a theoretical point of view, although the first con-
recently attracts the interest of chemists due to its simplicity fined level is usually elevated, resulting in an increase of the
of wet-chemistry synthesis. Recent experiments and calcula- absorption edge, one needs to note that Tauc plot is in fact
tions both confirm it has an optical band gap of around an extrapolation technique that utilizes the trend of absorp-
1.0 eV.13–15 Fig. 3(d) clearly shows an overestimation of the tion coefficient in the quasi-continuous spectrum regime.
Tauc fitted band gap, if 1 is used as the Tauc factor. The Therefore, it relates to the state distribution, where the dis-
underestimation with a Tauc factor of 1/2 [Fig. 3(c)] is rela- crete energy spectrum becomes quasi-continuous. With
tively not so convincing. But this causes a significant degra- appropriate smoothing, the general trend of density of states
dation of its linearity, as the measured absorption curves in nanocrystals does not vary from their bulk counterparts,
show obvious convex features. The band gap values fitted by instead the momentum selection rule of optical transitions is
Tauc plot with different power factors are listed In Table I partly relaxed. This argument supports the point that Tauc
for both NC materials. The optimized power factors result in plot could only be used to extrapolate the bulk band gap for
values that perfectly match the confirmed optical band gaps. direct-band-gap materials.
Interestingly, the Tauc plots used here retrieve the opti- To further illustrate this point, we consider the effect of
cal band gap of the respective bulk materials, and do not ex- carrier effective masses to the optical transition rate.
plicitly demonstrate blue shift attributed by quantum According to (29), the general form of direct-allowed optical
confinement. In fact, despite the fact that quantum confine- transition rate only depends on the ratio of the two effective
ment leads to blue shift on the absorption edge, using Tauc masses (i.e., me =mh ) except for a constant factor. This
plot extrapolation to determine the elevated bandgap value is implies that if we increase the two effective masses simulta-
completely misleading. According to Figs. 2(c) and 3(c), the neously while conserving their ratio, the relation between the
extrapolated energy is well above the band gap energy of the excess photon energy and the transition rate does not vary;
while in this case, the first confined level red shifts. This sup-
TABLE I. Comparison of Tauc plots with different values of the factor n. ports the argument that Tauc extrapolation is not capable of
finding the first confined energy level.
Tauc fitted band gaps (in unit of eV)
V. CONCLUSION
NC
material Optimal power factor Optimal n ¼ 1/2 n¼1 In summary, we have formulated the direct-allowed opti-
SnO2 0.80 (3.6 nm) – 0.82 (3.0 nm) 3.9 4.3 3.7
cal transition in semiconductor nanocrystals, as well as provid-
Ag2S 0.64 (5.3 nm) – 0.66 (6.8 nm) 0.99 1.03 0.88 ing a solid theoretical support of using Tauc plot extrapolation
with a proper power factor. Observed in this work, the power

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52
 125701-7 Feng et al. J. Appl. Phys. 117, 125701 (2015)

5
factor undergoes a size-dependent transition from one half in J. I. Pankove, Optical Processes in Semiconductors (Courier Dover
bulk semiconductors to one in ultra-small particles. By adopt- Publications, 2012).
6
P. Ardalan, T. P. Brennan, H.-B.-R. Lee, J. R. Bakke, I.-K. Ding, M. D.
ing the size-dependent power factor in extrapolation, a better McGehee, and S. F. Bent, “Effects of self-assembled monolayers on solid-
linearity and more consistent and accuracy bandgap value can state CdS quantum dot sensitized solar cells,” ACS Nano 5, 1495–1504
be retrieved, which allows us to characterize the material com- (2011).
7
P. Praus, O. Kozak, K. Kočı, A. Panaček, and R. Dvorsk
y, “CdS nanopar-
position of nanocrystals more accurately. However, for direct-
ticles deposited on montmorillonite: Preparation, characterization and
band-gap materials, the Tauc plot technique should not be used application for photoreduction of carbon dioxide,” J. Colloid Interface Sci.
to quantitatively determine the blueshift of bandgap value due 360, 574–579 (2011).
8
to quantum confinement effect. S. Park, E. Cho, D. Song, G. Conibeer, and M. A. Green, “n-Type silicon
quantum dots and p-type crystalline silicon heteroface solar cells,” Sol.
Energy Mater. Sol. Cells 93, 684–690 (2009).
ACKNOWLEDGMENTS 9
J.-W. Lee, D.-Y. Son, T. K. Ahn, H.-W. Shin, I. Y. Kim, S.-J. Hwang, M.
J. Ko, S. Sul, H. Han, and N.-G. Park, “Quantum-dot-sensitized solar cell
This program has been supported by the Australian
with unprecedentedly high photocurrent,” Sci. Rep. 3, 1050 (2013).
Government through the Australian Renewable Energy 10
L. D. Landau and L. M. Lifshitz, Quantum Mechanics Non-Relativistic
Agency (ARENA). Responsibility for the views, Theory, 3rd ed. (Butterworth-Heinemann, 1981), Vol. 3.
11
information, or advice expressed herein is not accepted by T. L. Chow, Mathematical Methods for Physicists (Cambridge University,
United Kingdom, 2000).
the Australian Government. This work was also supported 12
H. Tada, A. Hattori, Y. Tokihisa, K. Imai, N. Tohge, and S. Ito, “A pat-
by NSW Government through the Science Leverage Fund. terned-TiO2/SnO2 bilayer type photocatalyst,” J. Phys. Chem. B 104,
4585–4587 (2000).
1 13
V. Srikant and D. R. Clarke, “On the optical band gap of zinc oxide,” T. G. Schaaff and A. J. Rodinone, “Preparation and characterization of sil-
J. Appl. Phys. 83, 5447–5451 (1998). ver sulfide nanocrystals generated from silver (i)-thiolate polymers,”
2
T. Tansley and C. Foley, “Optical band gap of indium nitride,” J. Appl. J. Phys. Chem. B 107, 10416–10422 (2003).
14
Phys. 59, 3241–3244 (1986). M. C. Brelle, J. Z. Zhang, L. Nguyen, and R. K. Mehra, “Synthesis and
3
W. Yim, E. Stofko, P. Zanzucchi, J. Pankove, M. Ettenberg, and S. ultrafast study of cysteine- and glutathione-capped Ag2S semiconductor
Gilbert, “Epitaxially grown AlN and its optical band gap,” J. Appl. Phys. colloidal nanoparticles,” J. Phys. Chem. A 103, 10194–10201 (1999).
15
44, 292–296 (1973). R. Vogel, P. Hoyer, and H. Weller, “Quantum-sized PbS, CdS, Ag2S,
4
J. Tauc, “Optical properties and electronic structure of amorphous Ge and Sb2S3, and Bi2S3 particles as sensitizers for various nanoporous wide-
Si,” Mater. Res. Bull. 3, 37–46 (1968). bandgap semiconductors,” J. Phys. Chem. 98, 3183–3188 (1994).

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
138.5.159.110 On: Wed, 25 Mar 2015 06:30:52