See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/250928866

Reassessment of the Limiting Efficiency for Crystalline Silicon Solar Cells

Article in IEEE Journal of Photovoltaics · July 2013

DOI: 10.1109/JPHOTOV.2013.2270351

CITATIONS READS

800 5,281

3 authors:

Armin Richter Martin Hermle

Fraunhofer Institute for Solar Energy Systems ISE Fraunhofer Institute for Solar Energy Systems ISE
105 PUBLICATIONS 5,005 CITATIONS 385 PUBLICATIONS 13,651 CITATIONS

SEE PROFILE SEE PROFILE

Stefan W. Glunz
University of Freiburg
590 PUBLICATIONS 22,012 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Internal Resistance of PERT solar cells View project

DISC (Double Side Contacted Cells) - a H2020 project View project

All content following this page was uploaded by Stefan W. Glunz on 17 May 2014.

The user has requested enhancement of the downloaded file.

Accepted for IEEE Journal of Photovoltaics 2013 1

Reassessment of the Limiting Efficiency

for Crystalline Silicon Solar Cells
Armin Richter, Martin Hermle, and Stefan W. Glunz

 holes [5]. Kerr et al. addressed the effect of this Coulomb-

Abstract—Recently, several parameters relevant for modeling enhanced Auger recombination on the limiting efficiency [6]
crystalline silicon solar cells were improved or revised, e.g. the by applying their general parameterization of the intrinsic
international standard solar spectrum or properties of silicon recombination rate [7], which results in a maximum efficiency
such as the intrinsic recombination rate and the intrinsic carrier
concentration. In this work we analyzed the influence of these
of 29.05% for solar cells made of undoped silicon.
improved state-of-the-art parameters on the limiting efficiency Subsequently, the quantitative description of the radiative
for crystalline silicon solar cells under one-sun illumination at and the Auger recombination were improved further. Trupke
25 °C, by following the narrow-base approximation to model et al. measured thoroughly the radiative recombination
ideal solar cells. We considered also bandgap narrowing, which coefficient B [8], which appeared to be 50% smaller than the
was not addressed so far with respect to efficiency limitation. The value used by Kerr et al. [7, 6], and Altermatt et al. extended
new calculations presented in this work result in a maximum
theoretical efficiency of 29.43% for a 110 µm thick solar cell
the radiative recombination model by Coulomb interactions
made of undoped silicon. A systematic calculation of the I-V [9] similarly to Coulomb-enhanced Auger recombination. The
parameters as a function of the doping concentration and the cell Auger recombination according to the most commonly used
thickness together with an analysis of the loss current at parameterization of Kerr and Cuevas [7] was found to be
maximum power point provide further insight into the intrinsic significantly overestimated for lowly doped silicon
limitations of silicon solar cells. (<1016 cm3) [10], which is relevant for the optimum c-Si solar
cell. Taking the Coulomb-enhanced radiative recombination
Index Terms—Efficiency limit, silicon, solar cells
into account, an improved parameterization for the intrinsic
I. INTRODUCTION bulk lifetime was introduced in Ref. [10].
Besides these intrinsic recombination properties, also other
T HE maximum power conversion efficiency of silicon solar
cells is limited by intrinsic properties of silicon, such as
the band gap energy or the charge carrier recombination
parameters relevant for the efficiency limit of c-Si solar cells
were recently improved: (i) a new internationally accepted
reference solar spectrum was introduced [11] which led to a
properties. This intrinsic limitation is of basic interest for PV significant increase of revised experimental solar cell peak
research, particularly as continuous improvements in efficiencies [12], (ii) Green presented self-consistent optical
technology lead to solar cell operation conditions approaching parameters of intrinsic silicon based on improved
this fundamental limit, e.g. the open-circuit voltage of hetero- experimental data [13], and (iii) Rüdiger et al. determined new
junction silicon solar cells [1]. parameters for FCA explicitly for a wavelength range below
Based on detailed balancing of incident and generated 2 µm [14], the relevant range for c-Si solar cells.
power density, Shockley and Queisser determined an upper Further, Mattheis et al. showed that the charge carrier
efficiency limit for ideal single junction solar cells taking only mobility cause no inherent limitation to the maximum
radiative recombination into account [2]. In crystalline silicon, efficiency as it is the case e.g. for solar cells with amorphous
however, there are also considerable non-radiative intrinsic silicon base [15]. The influence of bandgap narrowing (BGN)
loss processes, in particular Auger recombination or parasitic on the limiting efficiency, however, was not addressed so far.
free-carrier absorption (FCA). By calculating the limiting BGN affects the effective intrinsic carrier concentration ni,eff,
efficiency including non-radiative losses, both Green [3] and which increases considerably with increasing charge carrier
Tiedje et al. [4] showed that actually Auger recombination is concentration, in particular also with increasing injection level
the dominant intrinsic loss mechanism for c-Si solar cells at [16, 17] (cf. Fig. 1). The I-V characteristic of silicon solar
one-sun illumination. Both even underestimated the Auger 2
cells is a function of n i,eff , and thus depends sensitively on
recombination as they used the traditional free-particle model,
which accounts not for the increased recombination ni,eff. Therefore, BGN becomes not only important for high
probability caused by Coulomb-interactions of electron and base doping concentrations, but also for low base doping
concentration under high injection [18], as it is the case for the
Manuscript received XX, 2013. This work was supported by the German
optimum silicon solar cell.
Federal Ministry for the Environment, Nature Conservation and Nuclear In this paper, we calculated the limiting efficiency for single
Safety under contract number 0325292‘‘ForTeS’’. junction silicon solar cells under one-sun illumination
The authors are with the Fraunhofer Institute for Solar Energy Systems
(ISE), 79110 Freiburg, Germany (e-mail: [email protected]).
(AM1.5G) at 25 °C based on state-of-the-art modeling
Accepted for IEEE Journal of Photovoltaics 2013 2

parameters. In doing so, we address the effect on the where JL is the current density generated by the incident light,
efficiency limit for each parameter separately, which differs W the cell thickness, Rintr the intrinsic recombination rate
from the frequently cited calculations of Kerr et al. [6]. These (radiative and Auger recombination), and q the elementary
are (i) the new solar spectrum, (ii) the updated optical charge. The intrinsic bulk lifetime of silicon τintr and
properties of silicon, (iii) the new parameters for FCA, (iv) the consequently also Rintr = Δn/τintr are an explicit function of the
improved description of radiative and Auger recombination, excess carrier concentration Δn. Under the assumption of a
and (v) the effect of BGN. Finally, we investigate in detail the “narrow base” the change of the quasi-Fermi levels within the
limiting efficiency as a function of the base doping base are very small, and the quasi-Fermi level separation can
concentration and the wafer thickness. be considered to be constant [3]. When assuming additionally
ideal contacts, the quasi-Fermi level separation equals the cells
output voltage V, and thus, Δn is related to V by [3, 4]:

 qV 
np   n 0   n  p 0   n   n i,eff
2
exp   (2)
 k BT 

with the electron and hole concentration n and p, the

respective equilibrium concentration n0 and p0, the Boltzmann
constant kB and the temperature T. This relation is used to
determine Δn(V, n0, p0) for Rintr(n0, p0, Δn) in Eq. (1).
Actually, Eq. (1) is also based on the narrow base assumption
as it implies a constant recombination rate throughout the cell,
which is fulfilled if the quasi-Fermi levels are constant. It
should be mentioned that by modeling the I-V characteristic
with Eq. (1), we assume in addition to ideal contacts also
negligible series resistance losses arising from the charge
carrier transport within the silicon bulk. Obviously, this is a
good assumption for cells made of highly doped silicon.
However, it is also a good assumption for cells made of lowly
doped silicon, as these are in sufficient high injection at
maximum power point (cf. Sec. III.B).
B. Photo-generation
The absorbed current density JL was calculated assuming
that every photon absorption event generates a single electron-
hole pair. The creation of multiple electron-hole pairs by high
energy photons [21] was neglected, as it is quite a small
effected contributing less than 0.1 mA/cm2 [22, 20]. Further, a
Lambertian randomizing light-trapping scheme together with
Fig. 1. Effective intrinsic carrier concentration ni,eff for n-type Si (upper an isotropic response of the cell was assumed, which results in
graph) and p-type Si (lower graph) at 25 °C as a function of Δn for different
base doping concentrations. ni,eff is calculated according to Refs. [16, 17, 19] an increased mean path length for a light ray inside the cell of
as described in Sec. II.D. 4nr2W , where nr is the refractive index of silicon [4]. The
weakly absorbed sub-bandgap photons have also a certain
probability of being absorbed by free-carriers. Taking this
II. MODELING OF THE I-V CHARACTERISTIC
parasitic FCA into account, JL can be determined from:
A. General Approach

To determine the maximum efficiency of silicon solar cells J L  q  Abb ( E ) Φ ( E ) dE (3)
limited by intrinsic properties of silicon, we follow the 0

approach of previous publications in modeling ideal cells

without surface and defect recombination, as well as perfect with the spectrally resolved relative absorbance in the cell [4]:
front side antireflection coatings and perfect reflecting rear
 bb ( E )
mirrors [4, 20, 6]. In doing so, we do not consider any Abb ( E )  ,
1 (4)
efficiency enhancement associated with impurity bands,  bb ( E )   FC A ( E ) 
improved internal confinement due to quantum structures, or 4 n r2W
spectrum-shifting (up or down conversion). The current-
voltage characteristic of such an ideal cell can be expressed as: the solar spectrum Φ normalized to an intensity of 0.1 W/cm2,
and the absorption coefficients for band-to-band transitions
J  J L  qWR intr (1) and FCA, αbb and αFCA, both as a function of the photon
energy E. Actually, Abb(E) is also a function of electron and
hole concentration, as αFCA(E) is proportional to n and p. The
last term in the denominator of Eq. (4) is the external emission
Accepted for IEEE Journal of Photovoltaics 2013 3

rate, which is strictly valid only in the weakly absorbing limit; Fig. 2. Spectrally resolved probability for photon-recycling PPR (band-to-
band transitions), free-carrier absorption PFCA, and external emission Pex of
however, as Abb(E) is close to unity in the strong absorbing radiatively emitted photons for different injection levels Δn. The probabilities
range, Eq. (4) can be used for the whole energy range [4]. were calculated for a 200-µm-thick undoped silicon wafer at 25 °C.
Unless stated otherwise, the solar spectrum of Ref. [11] was Additionally shown is the normalized spectrally resolved spontaneous
used, FCA according to Ref. [14], and tabulated data for nr emission B(E). The external emission was calculated assuming Lambertian
light trapping. The arrows show the increasing contribution of FCA with
and αbb from Ref. [13]. Particularly in section III.A, the increasing Δn and its effect on PPR and Pesc. Eg,0 shows the silicon bandgap
influence of differing parameters will be discussed. without BGN.
C. Intrinsic Recombination and Photon Recycling
In section III.A we also compare the parameterizations of D. Intrinsic Carrier Concentration
the intrinsic recombination rate of Kerr et al., Rintr,K, (Eq. (24) The effective intrinsic carrier concentration ni,eff is
in [7]) with the recently improved description of Richter et al., calculated from the expression [17]:
Rintr,R (Eq. (18) in [10]). Both account for radiative
recombination. Each radiative recombination event generates a  Eg 
photon, which has actually a high probability of being n i,eff  ni,0 exp   (5)
reabsorbed by band-to-band transitions in case of the very  2 k BT 
effective light-trapping assumed here [4, 6]. This reabsorption with the intrinsic carrier concentration ni,0, and the bandgap
is also referred to as “photon-recycling” (PR). To account for narrowing ΔEg according to Schenk [16]. ni,eff is calculated
PR, the radiative recombination coefficient B in Rintr is using ni,0 for Si at 300 K of Altermatt et al. [17]. As all
multiplied with (1  PPR), where PPR is the PR probability [6]. calculations of this work were performed for 25 °C, the
Following Refs. [4, 23], we determined PPR analytically
standard testing conditions of silicon solar cells, ni,0 was
assuming Lambertian light-trapping, as included in Eq. (4),
scaled using the temperature dependence of Sproul and Green
also for the luminescent photons to be consistent with the
treatment of the incident photons. The resulting PPR is in good [26], as suggested in Ref. [19]. This results in ni,0 =
agreement with Ref. [24], but differs from the ray tracing 8.28109 cm3 at 25 °C. Fig. 1 shows ni,eff as a function of Δn
approach of Kerr et al. [6], which results, for instance, in an for n-type and p-type Si with different doping concentrations,
12% higher PPR for W = 100 µm (cf. Fig. 6 in the Appendix). illustrating the significant increase at high excess carrier
The spectral dependence of the spontaneous emission B(E), as densities.
required to determine PPR, was calculated from the generalized E. Calculation of I-V Parameters
Planck equation, which is an extended form of the van
Roosbroeck equation [25, 8]. Fig. 2 shows the spectrally Finally, the efficiency was calculated numerically by
resolved probability for photon-recycling PPR (band-to-band solving Eq. (1) iteratively for maximum power point
transitions), free-carrier absorption PFCA, and external conditions, i.e. when d(JV)/dV equals zero. Therefore, Eq. (2)
emission Pex (calculated from Eq. (8) in the Appendix), all for is expressed as Δn(V), to calculate Rintr(n0, p0, Δn(V)) and
200 µm thick undoped Si at different injection levels. JL(n0, p0, Δn(V)) in Eq. (1). To determine the open-circuit
Throughout this work, “undoped” refers to calculations voltage VOC, Eq. (1) was similarly solved for J equals zero.
assuming an ultra-low doping concentration of 1011 cm3. Due The fill factor FF was calculated from the maximum power,
to their nature, Pex dominates the weakly absorbed sub- VOC and the short-circuit current density JSC.
bandgap photons, while PPR dominates at high photon
energies. For injection levels Δn > 1016 cm3, FCA becomes III. RESULTS AND DISCUSSION
significant in the sub-bandgap range. Although, FCA mainly A. Influence of Solar Spectrum and Silicon Parameters
affects Pex, it reduces also PPR, particularly under open-circuit In this section we address the effect of following improved
conditions due to the high injection levels (cf. Sec. III.B). modeling parameters on the limiting efficiency: (i) new solar
spectrum, (ii) updated optical properties of silicon, (iii) new
parameters for FCA, (iv) improved description of radiative
and Auger recombination, and (v) the effect of BGN. We
discuss these effects successively. As a starting point, we used
a modeling parameter configuration similar to that of Kerr et
al. [6] with two exceptions: the photon recycling was
calculated analytically instead of the ray-tracing approach of
Kerr et al. (cf. Sec. II.C and Appendix), and ni,eff = ni,0 =
8.28109 cm3 instead of the 8.65109 cm3, to compare the
effect of BGN within one model (cf. Sec. II.D). The effects are
discussed for undoped silicon, as the maximum efficiency is
reached in the limit of zero doping [4, 6]. The resulting I-V
parameters of this initial modeling configuration are shown in
row A of Tab. I.
Accepted for IEEE Journal of Photovoltaics 2013 4

TABLE I: I-V PARAMETERS AND OPERATION CONDITIONS OF OPTIMUM SILICON SOLAR CELLS UNDER ONE-SUN ILLUMINATION AT 25 °C. THE CALCULATION
RESULTS ARE COMPARED FOR DIFFERENT SETS OF MODELING PARAMETERS. ALL RESULTING OPERATION PARAMETERS ARE SPECIFIED IN THE CAPTION OF
FIG. 4. THE RESPECTIVE OPTIMUM CELL THICKNESS Wopt IS ALSO GIVEN.
Modification Modeling parameters η VOC JSC FF Vmpp Δnmpp ΔnOC Wopt
a n r, 3 3
Φ FCA c Rintr d ni,eff e
(%) (mV) (mA/cm ) 2
(%) (mV) (cm ) (cm ) (µm)
αbb b
15 16
A Φ1st old αFCA,G Rintr,K const. 29.12 769.7 42.75 88.49 700.5 6.8810 2.6510 85
B spectrum Φ2nd old αFCA,G Rintr,K const. 29.32 767.4 43.19 88.47 698.2 6.581015 2.531016 95
C optical properties Φ2nd new αFCA,G Rintr,K const. 29.30 767.3 43.16 88.46 698.2 6.581015 2.531016 95
15 16
D FCA Φ2nd new αFCA,R Rintr,K const. 29.30 767.4 43.16 88.47 698.2 6.5910 2.5310 95
E recombination Φ2nd new αFCA,R Rintr,R const. 29.57 767.3 43.31 88.98 700.9 6.951015 2.531016 110
F bandgap narrowing Φ2nd new αFCA,R Rintr,R BGN 29.43 761.3 43.31 89.26 697.3 6.901015 2.531016 110
a
solar spectrum: Φ1st according to Ref. [27]; Φ2nd according to Ref. [11].
b
refractive index nr and absorption coefficient αbb of silicon: “old” according Ref. [28]; “new” according to Ref. [13].
c
free-carrier absorption (FCA): αFCA,G according to Green [22]; αFCA,R according to Rüdiger et al. [14].
d
intrinsic recombination rate of silicon (radiative and Auger recombination): Rintr,K according to Kerr and Cuevas [7]; Rintr,R according to Richter et al. [10].
e
effective intrinsic carrier concentration: “const.” without BGN, i.e. ni,eff = ni,0 = 8.28109 cm3; “BGN” with BGN according to Schenk [16].

In a first step, the old international standard solar spectrum Altogether, row F represents the I-V parameters for the
AM1.5G [27] was replaced by the one introduced in 2008 optimum silicon solar cell calculated with state-of-the-art
[11]. As can be seen from row B of Tab. I, this increases the modeling parameters. The resulting efficiency limit is 29.43%,
maximum efficiency by 0.20%-abs., mainly due to a 0.38%-abs. higher than the 29.05% calculated by Kerr et al.
0.44 mA/cm2 increase of JSC. This JSC increase is in good [6]. Fig. 3 shows the limiting efficiency calculated with these
agreement with the 0.5 mA/cm2 increase observed for modeling parameters as a function of cell thickness for
experimental silicon solar cell results revised with the new undoped Si, as discussed so far, as well as for 1 Ω cm p-type
spectrum [12]. Si. The 1 Ω cm p-type Si calculation of Kerr et al. (Fig. 2 of
Second, the optical properties of silicon [28] were replaced Ref. [6]) is also included. The latter results in a peak
by the more recently published self-consistent data set of efficiency of 28.6%, which is 0.54%-abs. lower than the
Green [13]. By comparing row C with the previous row B, no 29.14% of this work. Thus, the difference is even more
considerable effect on the efficiency (+0.02%-abs.) is pronounced for higher doping concentration.
observed. This results from the fact that there is no difference
in the refractive index nr between both, and only a difference
of < 5% for αbb in the weakly absorption range, the most
relevant range for determining JL.
In row D, the FCA parameterization of Green [22] was
replaced by the recent parameterization of Rüdiger et al. [14].
Again, no considerable effect on the efficiency is observed.
Even if there is a certain difference between both, FCA is only
a second order effect, which explains the marginal influence
(cf. Sec. III.B and Fig. 5).
Next, the parameterization of the intrinsic recombination
rate Rintr,K [7] is replaced by the recent parameterization Rintr,R
[10]. As can be seen from row E, Rintr,R results in a significant
higher efficiency (0.27%-abs.) than Rintr,K, which originates
from significant lower recombination rates for radiative
recombination [8, 9] as well as Auger recombination [10].
Interestingly, this results not in a higher VOC, but in a higher
FF, due to the higher maximum power point voltage Vmpp, and
a higher JSC. The reduced recombination rate leads to an
increase of the optimal cell thickness by ~15 µm, which is the Fig. 3. Efficiency as a function of the thickness for silicon solar cells under
reason for the higher JSC. 1-sun illumination and at 25 °C. The solid lines represent the efficiency for a
Finally, the results of row F were calculated taking BGN for cell made of 1 Ω cm p-type Si and undoped Si, taking radiative and Auger
recombination as well as photon recycling (PR) into account. The additional
the determination of ni,eff into account, instead of assuming curves show cells constrained either only by Auger recombination or by
ni,eff = ni,0 to be constant at 8.28109 cm3. This results in a Auger and radiative recombination. Both are calculated without taking PR
efficiency reduction off 0.14%-abs., which is mainly due to a into account and thus, represent an upper limit assuming complete photon
recycling, and lower limit assuming no photon recycling, respectively. All
6 mV lower VOC level and consequently also a lower Vmpp
these curves were calculated with the modeling parameters listed in row F of
level, while the FF even increased slightly. The difference in Tab. I. Additionally, the curves are also shown for 1 Ω cm p-type Si
Vmpp of 3.6 mV is in good agreement with the 3.4 mV calculated with the modeling parameters used by Kerr et al. [6]. The cross
originally estimated by Tiedje et al. [4] for neglecting BGN symbols are the respective data points digitized from Fig. 2 of Ref. [6].
according to Lanyon and Tuft [29].
Accepted for IEEE Journal of Photovoltaics 2013 5

Fig. 4. Limiting efficiency and I-V parameters of n-type Si (upper graphs) and p-type Si (lower graphs) solar cells as a function of the cell thickness and the
doping concentration. The results are calculated for cells under 1-sun illumination (AM1.5G) at 25 °C, which are constrained by radiative and Auger
recombination taking photon recycling and FCA into account. The modeling parameters of row F in Tab. I were used for these calculations. In addition to the
efficiency η, also shown are the open-circuit voltage VOC, the voltage at maximum power point Vmpp, the short-circuit current density JSC, the fill factor FF, the
effective lifetime at mpp τmpp, as well as the injection levels at mpp and OC, Δnmpp and ΔnOC, respectively. The ratio of the minority carrier diffusion length L and
the thickness, L/W, is < 2 in the white upper right corner of each plot, and < 5 right to the dashed line. As discussed in the text, this both range have limited
accuracy due to the narrow-base approximation.

Fig. 3 shows also upper and lower bounds for the luminescence light of the radiative recombination, the silicon
efficiency, which were calculated assuming complete and no absorbance and consequently the photon recycling increases
photon recycling, respectively. For the weakly absorbed considerable with increasing cell thickness (cf. Fig. 6 in the
Accepted for IEEE Journal of Photovoltaics 2013 6

appendix). For that reason, the resulting efficiency curve with the low- and high-injection contribution, JA,li and JA,hi
photon recycling is close to the lower bound for thin cells, and respectively, as well as the effective radiative recombination
converge to the upper bound for thick cells. For 1 Ω cm p-type Jrad,eff, which is reduced by the fraction of recycled photons.
Si, the calculation of Kerr et al. results in a higher difference The current density of the recycled photons Jrad,PR is shown too
between both bounds (35% higher at 80 µm), which directly as dashed area above Jloss,mpp.
reflects the higher radiative recombination rate included in The Auger recombination clearly dominants the whole Ndop
Rintr,K. range particularly as a distinct part of the radiative
recombining charge carriers are recycled. While at low doping
B. Influence of Base Doping Concentration
concentrations JA,hi dominants, JA,li dominates at high doping
The results discussed in the previous section reveal a concentrations, which directly reflects the transition from high
significant influence of Rintr and ni,eff on the limiting efficiency, injection to low injection at maximum power point (cf. Δnmpp
both strongly affected by the base doping concentration Ndop. in Fig. 4). Although there is a small contribution of FCA for
Thus, the limiting efficiency will also vary considerably with Ndop > 1016 cm3, the main increase of Jloss,mpp with increasing
Ndop. Fig. 4 shows the limiting efficiency as well as the I-V Ndop is caused by the absolute increasing contribution of Auger
parameters for n-type and p-type Si cells as a function of Ndop recombination.
and W, using the state-of-the-art modeling parameters as listed
in row F of Tab. I, particularly taking also BGN into account.
It is important to note that the narrow-base approximation
used throughout this work to relate the cell voltage with Δn
according to Eq. (2) is strictly valid only for W << L, with L
being the minority carrier diffusion length [3]. If L is in the
same order as W or even smaller, non-uniform charge carrier
profiles lead to significant gradients in the minority carrier
quasi-Fermi levels, which affects the cell voltage. The device
simulator PC1D [30] was used to estimate roughly the
influence of minority carrier quasi-Fermi level gradients on
the cell voltage. According to this simulation, the effect on the
voltage is < 2.5 mV for L/W > 2, and < 0.3 mV for L/W > 5.1
These ranges are also shown in Fig. 4: L/W is < 2 in the white
upper right corner of each plot, and < 5 right to the dashed
line. Note that the results at the maximum power point (mpp)
are less influenced than at open-circuit conditions, due to the
higher effective lifetime at mpp, τmpp, which is associated with
the lower injection level at mpp, Δnmpp. In particular, for the
absolute limiting efficiency calculation discussed in Sec. III.A,
Fig. 5. Loss current density at maximum power point Jloss,mpp (thick black
the narrow-base assumption is well valid [3, 4].
line, upper graph) as a function of the doping concentration Ndop for ideal
As expected, the limiting efficiency η for n-type and p-type n-type Si solar cells with a thickness of 110 µm. Shown are the contribution
Si cells shown in Fig. 4 reach the maximum in the limit of of the incident photons absorbed by free-carriers JFCA, the recombination
undoped Si with a thickness of ~110 µm. For lower current of Auger recombination, separated into the low- and high-injection
contribution, JA,li and JA,hi respectively, as well as the effective radiative
thicknesses, η increases with increasing W due to improved
recombination Jrad,eff, which is reduced by the fraction of recycled photons.
light absorption quantified by JSC. Although Δnmpp decreases The current density of this recycled photons Jrad,PR is shown too as dashed
with increasing W, which results in an increasing τmpp, Vmpp area above Jloss,mpp. The lower graph shows the local ideality factor mmpp
and also VOC decrease with increasing W due to an increasing evaluated at mpp.
total recombination current density qWRintr in Eq. (1). This
increasing total recombination compensates the benefit of
increasing JSC at high thicknesses, and consequently decreases From Fig. 4 it can be seen that the fill factor FF decreases
η again. Rintr also increases with increasing Ndop, particularly with increasing Ndop, in case of the n-type cells significantly
above ~1015 cm3 in case of n-type Si and above ~1016 cm3 in above 1015 cm3. If series resistances are negligible, as
case of p-type Si. This leads to a reduction of Vmpp and VOC assumed, the FF is mainly influenced by the VOC and the cell
with increasing Ndop, which is more pronounced for n-type Si ideality factor m [32]. To some extent, therefore, the FF
mainly due to the different recombination kinetics of the low reduction can be explained by the VOC reduction at high
injection Auger recombination processes. doping concentrations. However, it is well known that m can
To illustrate the influence of the doping concentration, Fig. vary significantly with the doping and injection level.
5 shows the loss current density at maximum power point, According to the free-particle Auger model, the ideality factor
Jloss,mpp, as a function of Ndop for n-type Si cells with a of cells limited by Auger recombination is reduced from unity
thickness of 110 µm. Jloss,mpp is separated into the contributions under low injection conditions to 2/3 under high injection
of the incident photons absorbed by free-carriers JFCA, the conditions, which results in a significant increase of the upper
recombination current of Auger recombination, separated into FF bound [3]. To analyze the ideality factor for the cell
simulations of this work, we fitted the one-diode equation
1
(6)
L was calculated using the mobility model of Klaassen [31].
Accepted for IEEE Journal of Photovoltaics 2013 7

 [4] T. Tiedje, E. Yablonovitch, G. D. Cody, and B. G. Brooks, "Limiting

 V   k BT
J  J L  J 0  exp    1 with VT  efficiency of silicon solar cells", IEEE Trans. Electron Devices, vol. 31,
  mV T   q pp. 711-6, 1984.
[5] A. Hangleiter and R. Häcker, "Enhancement of band-to-band Auger
recombination by electron-hole correlations", Phys. Rev. Lett., vol. 65, pp.
to the modeled I-V characteristics in a voltage range around 215-18, 1990.
Vmpp by varying m and the recombination pre-factor J0. This [6] M. J. Kerr, A. Cuevas, and P. Campbell, "Limiting efficiency of crystalline
yields the local ideality factor at maximum power point mmpp. silicon solar cells due to Coulomb-enhanced Auger recombination", Prog.
The resulting dependency of mmpp on Ndop is shown exemplary Photovolt: Res. Appl., vol. 11, pp. 97-104, 2003.
[7] M. J. Kerr and A. Cuevas, "General parameterization of Auger
for the 110-nm-thick n-type Si cells in the lower graph of Fig. recombination in crystalline silicon", J. Appl. Phys., vol. 91, pp. 2473,
5. As expected, mmpp directly correlates with the transition 2002.
from high to low injection quantified by the transition from [8] T. Trupke, M. A. Green, P. Würfel, P. P. Altermatt, A. Wang, J. Zhao, and
JA,hi to JA,li. Similar to the free-particle Auger model, mmpp R. Corkish, "Temperature dependence of the radiative recombination
coefficient of intrinsic crystalline silicon", J. Appl. Phys., vol. 94, pp.
converges to unity in case of low injection, and is with 0.70 4930-7, 2003.
close to 2/3 under high injection. Actually, a value slightly [9] P. P. Altermatt, F. Geelhaar, T. Trupke, X. Dai, A. Neisser, and E. Daub,
higher than 2/3 is expected, as the denominator of the factor "Injection dependence of spontaneous radiative recombination in
2/3 originates from the exponent of the charge carrier crystalline silicon: experimental verification and theoretical analysis",
Appl. Phys. Lett., vol. 88, pp. 261901-1-3, 2006.
dependence of the high injection Auger recombination, which [10] A. Richter, S. W. Glunz, F. Werner, J. Schmidt, and A. Cuevas,
is 2.92 in Rintr,R [10]. This shows that for cells in high injection "Improved quantitative description of Auger recombination in crystalline
the bound on the fill factor is increased similarly to the known silicon", Phys. Rev. B, vol. 86, pp. 165202, 2012.
increase of the free-particle Auger model. [11] International Electrotechnical Commission, "Photovoltaic devices - part
3: measurement principles for terrestrial photovoltaic (PV) solar devices
with reference spectral irradiance data," International Standard, IEC
IV. CONCLUSION 60904–3, 2nd ed., 2008.
[12] M. A. Green, K. Emery, Y. Hishikawa, and W. Warta, "Solar cell
This work presents new calculations of the limiting efficiency tables (version 33)", Prog. Photovolt: Res. Appl., vol. 17, pp.
efficiency for crystalline silicon solar cells (in the “narrow 85-94, 2009.
base” approximation) systematically as a function of the [13] M. A. Green, "Self-consistent optical parameters of intrinsic silicon at
doping concentration and the cell thickness taking recently 300 K including temperature coefficients", Sol. Energy Mater. Sol. Cells,
vol. 92, pp. 1305-10, 2008.
improved modeling parameters into account. A detailed [14] M. Rüdiger, J. Greulich, A. Richter, and M. Hermle, "Parameterization of
analysis revealed significant effects of the new international free carrier absorption in highly doped silicon", accepted for publication
standard solar spectrum and the improved quantitative in IEEE Trans. Electron Devices.
description of Coulomb-enhanced radiative and Auger [15] J. Mattheis, J. H. Werner, and U. Rau, "Finite mobility effects on the
radiative efficiency limit of pn-junction solar cells", Phys. Rev. B, vol. 77,
recombination, while improved optical properties of silicon pp. 085203, 2008.
and a new parameterization for free-carrier absorption (FCA) [16] A. Schenk, "Finite-temperature full random-phase approximation model
have only a minor influence. The limiting efficiency was also of band gap narrowing for silicon device simulation", J. Appl. Phys., vol.
observed to be affected by bandgap narrowing particularly due 84, pp. 3684-95, 1998.
[17] P. P. Altermatt, A. Schenk, F. Geelhaar, and G. Heiser, "Reassessment of
to the high-injection conditions at maximum power point. the intrinsic carrier density in crystalline silicon in view of band-gap
Applying the state-of-the-art modeling parameters results in a narrowing", J. Appl. Phys., vol. 93, pp. 1598-604, 2003.
maximum theoretical efficiency of 29.43% for a 110 µm thick [18] P. P. Altermatt, "Models for numerical device simulations of crystalline
cell made of undoped Si. An evaluation of the loss current silicon solar cells—a review", Journal of Computational Electronics, vol.
10, pp. 314-30, 2011.
density at maximum power point reconfirm a dominating [19] A. Wolf, D. Biro, J.-F. Nekarda, S. Stumpp, A. Kimmerle, S. Mack, and
effect of Auger recombination, while the FCA contribution R. Preu, "Comprehensive analytical model for locally contacted rear
vanishes for doping concentrations below 1016 cm3 and the surface passivated solar cells", J. Appl. Phys., vol. 108, pp. 1-13, 2010.
contribution of radiative recombination is rather small due to [20] M. A. Green, "Limiting efficiency of bulk and thin-film silicon solar cells
in the presence of surface recombination", Prog. Photovolt: Res. Appl.,
photon recycling. vol. 7, pp. 327-30, 1999.
[21] S. Kolodinski, J. H. Werner, T. Wittchen, and H. J. Queisser, "Quantum
ACKNOWLEDGMENT efficiencies exceeding unity due to impact ionization in silicon solar
cells", Appl. Phys. Lett., vol. 63, pp. 2405-7, 1993.
The authors thank Dr Patrick Campbell for valuable [22] M. A. Green, Silicon solar cells: Advanced principles and practice.
discussions about photon recycling. Sydney, NSW: Centre for Photovoltaic Devises and Systems UNSW,
1995.
[23] O. D. Miller, E. Yablonovitch, and S. R. Kurtz, "Strong internal and
REFERENCES external luminescence as solar cells approach the Shockley–Queisser
[1] K. Maki, D. Fujishima, H. Inoue, Y. Tsunomura, T. Asaumi, S. Taira, T. limit", IEEE Journal of Photovoltaics, vol. 2, pp. 303, 2012.
Kinoshita, M. Taguchi, H. Sakata, H. Kanno, and E. Maruyama, "High- [24] M. A. Green, J. Zhao, A. Wang, P. J. Reece, and M. Gal, "Efficient
efficiency HIT solar cells with a very thin structure enabling a high VOC", silicon light-emitting diodes", Nature, pp. 805-8, 2001.
in Proceedings of the 37th IEEE Photovoltaic Specialists Conference, [25] P. Würfel, "The chemical potential of radiation", Journal of Physics C:
Seattle, Washington, USA, 2011. Solid State Physics, vol. 15, pp. 3967-3985, 1982.
[2] W. Shockley and H. J. Queisser, "Detailed balance limit of efficiency of p- [26] A. B. Sproul and M. A. Green, "Intrinsic carrier concentration and
n junction solar cells", J. Appl. Phys., vol. 32, pp. 510-9, 1961. minority-carrier mobility of silicon from 77 to 300 K", J. Appl. Phys., vol.
[3] M. A. Green, "Limits on the open-circuit voltage and efficiency of silicon 73, pp. 1214-25, 1993.
solar cells imposed by intrinsic Auger processes", IEEE Trans. Electron [27] International Electrotechnical Commission, "Photovoltaic devices - part
Devices, vol. 31, pp. 671-8, 1984. 3: measurement principles for terrestrial photovoltaic (PV) solar devices
with reference spectral irradiance data," International Standard, IEC
60904–3, 1st ed., 1989.
Accepted for IEEE Journal of Photovoltaics 2013 8

[28] M. A. Green and M. J. Keevers, "Optical properties of intrinsic silicon at

300 K", Prog. Photovolt: Res. Appl., vol. 3, pp. 189-92, 1995.
[29] H. P. D. Lanyon and R. A. Tuft, "Bandgap narrowing in moderately to
heavily doped siliocn", IEEE Trans. Electron Devices, vol. 26, pp. 1014,
1979.
[30] D. A. Clugston and P. A. Basore, "PC1D version 5: 32-bit solar cell
modeling on personal computers", in Proceedings of the 26th IEEE
Photovoltaic Specialists Conference, Anaheim, California, USA, 1997.
[31] D. B. M. Klaassen, "A unified mobility model for device simulation. I.
Model equations and concentration dependence", Solid-State Electron.,
vol. 35, pp. 953-9, 1992.
[32] M. A. Green, "Accuracy of analytical expressions for solar cell fill
factors", Solar Cells, vol. 7, pp. 337-340, 1982.

APPENDIX: DETERMINATION OF THE PHOTON RECYCLING

To determine the photon recycling probably PPR, the
spectrally resolved radiative recombination coefficient B(E) Fig. 6. Photon recycling probability PPR for undoped Si as a function of the
was calculated from the generalized Planck equation [25, 8]: wafer thickness for different injection levels Δn. PPR was calculated assuming
Lambertian light trapping according to Tiedje et al. [4], and taking parasitic
1  E  free-carrier absorption into account. For Δn = 1015 cm3, PPR was exemplary
B(E )  nr2 ( E ) E 2 bb ( E ) exp    (7) calculated also with Lambertian light trapping according to Green et al. [24].
 2 c02  3 ni,eff
2
 k BT  Also shown is PPR published by Kerr et al. determined with ray tracing
simulations [6].
with the photon energy E. PPR is then determined by
evaluating numerically:


PPR (W , n , p ) 
 0
Abb ( E , W , n , p ) B ( E ) dE
. (8)

0
B ( E ) dE

FCA is a function of n and p, and consequently also Abb given

in Eq. (4) and PPR. The probability for FCA or external
emission is calculated by replacing Abb by the respective term
analogously to the calculation of Abb in Eq. (4). It is worth
mentioning that the denominator of Eq. (8) results in B =
4.821015 cm3 s1 when using αbb(E) of Green et al. [13] in Eq.
(7). This value is in good agreement with the 4.731015 cm3 s1
determined by Trupke et al. [8], which is also included in
Rintr,R of Richter et al. [10], indicating consistency.
Fig. 6 shows PPR as a function the cell thickness W for
different injection levels Δn of undoped Si. The decreasing
PPR with increasing Δn is caused by to the increasing parasitic
FCA, as can be seen from Fig. 2. Below Δn = 1015 cm3, PPR
remains at the same level as for Δn = 1015 cm3, because FCA
is negligible for such low charge carrier concentrations (cf.
Fig. 2). Fig. 6 shows additionally PPR according to Green et al.
[24] (Lambertian light-trapping) exemplarily for
Δn = 1015 cm3, as well as PPR of the ray-tracing approach of
Kerr et al. (digitized from Fig. 1 of Ref. [6]). While the PPR
applied in this work is in good agreement with Green et al., it
is more conservative than the PPR of Kerr et al., particularly
around 100 µm, the thickness of the optimum Si solar cell.
Accepted for IEEE Journal of Photovoltaics 2013 9

View publication stats