Physics Unlimited Explorer Competition

Assignment packet

Guidelines
Student teams will have a total of up to two weeks to work on the 2023 Explorer Competition,
and teams registering later than the assignment release date may begin to work at any time.
For successful completion of this assignment, we recommend that teams set aside at least 10
hours of time, cumulatively. Please refer to the submission explanation below for details on
both formatting and the submission process.

Scoring
Students are encouraged to work on as much of the assignment as possible. The award structure
will be as follows:

1. Certificate awards will be given to the four teams with the highest scores, per scoring
rubric that would allocate points for questions and exercises based on their difficulty level
as determined by the assignment creator.

2. There will be medalist certificates granted for first place, second place, and third place,
respectively.

3. There will be an honorable mention certificate granted to the fourth-highest scoring sub-
mission.

Collaboration Policy and External Resources

Students participating in the competition may only correspond with members of their
team. Absolutely and unequivocally, no other form of human correspondence is allowed.
This includes any form of correspondence with mentors, teachers, professors, and other students.
Participating students are barred from posting content or asking questions related to the exam
on the internet (except where specified below), and moreover, they are unequivocally barred
from seeking the solution to any of the exams’ parts from the internet or another resource.
Students are allowed, however, to use the following resources for purposes of reference and
computation:

1
 • Internet: Teams may use the internet for purposes of reference with appropriate citation.
Again, teams are in no way allowed to seek the solution to any of the exams’ parts from
the internet. For information about appropriate citation, see below.
• Books and Other Literature: Teams may use books or other literature, in print or
online, for purposes of reference with appropriate citation. As with the use of the Internet,
teams are in no way allowed to seek the solution to any of the exams’ parts from books
or other literature.
• Computational Software: Teams may use computational software, e.g. Mathematica,
Matlab, Python, whenever they deem it appropriate. Of course, teams must clearly
indicate that they have used such software. Additionally, the judges reserve to right to
deduct points for the use of computational software where the solution may be obtained
simply otherwise.

Citation
All student submissions that include outside material must include numbered citations. We do
not prefer any style of citation in particular.

Submission
All submissions, regardless of formatting, should include a cover page listing the title of
the work, the date, and electronic or scanned signatures of all team participants.
The work must be submitted as a single PDF document with the “.pdf.” extension. All other
formatting decisions are delegated to the teams themselves. No one style is favored over another.
That being said, we recommend that teams use a typesetting language (e.g. LATEX) or a word-
processing program (e.g. Microsoft Word, Pages). Handwritten solutions are allowed, but
we reserve the right to refuse grading of any portion of a team’s submission in the
case that the writing or solution is illegible.

2
 Microcrystalline Organic Solar Cells

Contents
I Introduction 4

II Semiconductors 4
II.1 Semiconductors, Metals, and Insulators . . . . . . . . . . . . . . . . . . . . . . . 4
II.1.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
II.2 p vs. n-type Semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
II.2.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
II.3 Density of States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
II.3.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
II.4 Fermi Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
II.4.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
II.5 Electron and Hole Concentrations in Semiconductors . . . . . . . . . . . . . . . . 10
II.5.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
II.6 p-n Junction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
II.6.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
II.7 Current-Voltage Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
II.7.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

III Organic Semiconductors 17

III.1 Physics of Electron Arrangement in Atoms . . . . . . . . . . . . . . . . . . . . . 17
III.1.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
III.1.2 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
III.1.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
III.2 Molecular Bonding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
III.2.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
III.3 Energy Band Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
III.4 Absorption and Emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
III.4.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
III.4.2 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
III.4.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

IV Excitons 27
IV.1 Exciton Radii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
IV.1.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
IV.1.2 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
IV.1.3 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

V Data Analysis Primer 29

V.1 External Quantum Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
V.1.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

VI Data Analysis 30
VI.1 Extracting Exciton Energies from EQE . . . . . . . . . . . . . . . . . . . . . . . 30
VI.1.1 Exercise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Learning Goals
Through the course of this assignment, you will learn some basic concepts and ideas about
quantum mechanics and apply it to consider principles of designing a solar cell made of micro-

3
crystalline organic semiconductors. We want you to get an idea of the way fundamental physics
concepts form the foundations of modern, cutting-edge technologies, and for you to see that
even a little bit of physics knowledge can help you understand these devices a whole lot better.
Those of you who have worked on the 2018 Explorer assignment or looked over it while reviewing
might find some of the content familiar. However, rest assured each question will be different.

Topic Format and Grading

This document consists of a few sections of expository material with exercises and questions
interspersed along that lead you to considering several ideas related to laser design. Exercises
will typically ask students to mathematically derive or demonstrate a result useful to the dis-
cussion. They can also ask students to explain crucial concepts. Questions will ask students to,
in their own words, interpret stated results. We are looking to see how well you understand the
subject, so to receive full credit, all work shown must be complete and properly justified.

Expected Amount of Work

Do not expect to understand the concepts in this document after only one read through; these
concepts take time to absorb. While it may feel like you are not getting much accomplished
as you try to understand the reading, persevere. It may be necessary to read some passages
several times in a row before understanding them completely. Because there are not too many
questions in this document, you should have time to complete the readings. We have made
every attempt to be rigorous in our presentation, but simplifications have been made when
appropriate. Students are welcome to investigate the subject in more detail outside of this
document.

I Introduction
This exam assumes a familiarity with atomic structure and electron configuration. If you are
unfamiliar with these concepts, feel free to check out these Khan Academy resources on these
topics or Google the topics yourself to familiarze yourself with the necessary concepts.
The following sections present some background on semiconductors and solar cells to orient and
contextualize the exam and its content for you. Refer to the relevant sections of Chenming Hu’s
Modern Semiconductor Devices for Integrated Circuits or any other resource of choice for more
info or better explanations if needed. I would recommend Sections 1.3 and 1.4 for II, Sections
1.2, 1.7.1, 1.7.2, 4.1.1, and 4.1.2 for II.6,. You do not need to know all of the equations in these
sections unless specified in the exam, but you should understand the basic ideas and concepts
behind semiconductors versus metals and insulators, doping, Fermi levels, p-n junctions, thermal
equilibrium versus non-equilibrium in semiconductors, drift versus diffusion for carrier motion
within semiconductors, and the basic operating principle of a solar cell from a semiconductor
device perspective.

II Semiconductors
II.1 Semiconductors, Metals, and Insulators
When thinking about the electronic properties of materials, such as conductivity (which mea-
sures how well they conduct electricity), there are three main categories of materials:
• Metals: conduct electricity ”easily” (i.e. at room temperature, without adding impurities
to the material, needing to shine light on the material, etc.)

4
 • Insulators: conduct electricity poorly under most conditions
• Semiconductors: can be manipulated into conducting electricity by various means (eg.
adding impurities, shining light on them, etc.)
The difference, in practice, comes down to the electronic band structures of these materials.
When atoms form bonds with each other in a material, their orbitals overlap, and individual
orbitals become continuous bands within which electrons can reside. All shells with energies
at or below the valence shell form the valence band. The higher-energy orbitals, which are
empty when the atoms are in their lowest energy states, form the conduction band. The
conduction band is like a highway with no traffic, so if electrons can make their way into the
conduction band somehow (more on that later), then they can easily move. The valence band is
more like a Los Angeles highway during rush hour − electrons can barely move within it. The
conductivity of a material depends on the setup of these bands. Below is a comparison of the
band structures of metals, semiconductors, and insulators:

Figure 1: Comparison of the band diagram of metals, semiconductors, and insulators. The key difference lies in the energy
separation between the valence and conduction bands.

In metals, the valence bands and conduction bands overlap, so electrons can effortlessly move
from the valence to the conduction band, where they can conduct electricity when, say, a voltage
is applied to the material.
On the other end of the spectrum, an insulator, like glass, has such a large energy difference
between its conduction and valence states that it is practically impossible for electrons to jump
into the conduction state from the valence state under ”normal” circumstances. For example,
SiO2 has a bandgap of > 7 eV, which is in the UV part of the electromagnetic spectrum. When
heated to around 1340◦ F, or 730◦ C, glass can conduct electricity well enough to light a lightbulb.
This is because the high temperatures provide electrons with sufficient thermal energy to jump
into the conduction state in significant enough quantities to measure a current flow through the
device. The exact behavior of the electrons is probabilistic due to quantum mechanics, but on
a macroscopic level we can measure electric conduction even in insulators under such extreme
conditions.
Semiconductors are somewhere in between. Silicon, for instance, has a bandgap of 1.12 eV, which
is in the infrared part of the electromagnetic spectrum. This means that these materials can
conduct electricity more easily than insulators at elevated temperatures or under illumination.
An electron in silicon that absorbs a visible-light photon, for instance, can jump from the
valence to the conduction band. We can switch the material from being conductive to being

5
nonconductive with relatively quick modifications such as switching illumination or turning a
moderate voltage on and off. This property makes it incredibly useful for optoelectronic devices
such as lasers, LEDs, solar cells, and transistors. Modern electronic devices like your phone and
computer are made possible by this property of semiconductors!
Modern optoelectronic devices, however, generally do not use pure semiconductors to perform
this conductivity switching. Instead, they use something called p-n junctions, which will be
discussed more in the next section. These rely on deliberately inserting different impurities
in different parts of the semiconductor to make them conductive under certain voltages or
illumination levels.
Silicon is the most common semiconductor around. It is an inorganic semiconductor, and can
be crystalline, polycrystalline, or amorphous. Modern electronics are mostly made of crystalline
silicon, while solar cells can be made of all three types, although each has its advantages and
disadvantages as noted in the link above. Other inorganic semiconductors include gallium
arsenide (GaAs) and indium phosphide (InP), which are compound semiconductors made
of two elements rather than one. Semiconductors can get even more complicated, containing
three or more elements, such as indium aluminum arsenide (InAlAs). Organic semiconductors,
which will be the focus of this exam, are composed of long hydrocarbon molecules. Only some
of them can be crystallized. But they all can be used to manipulate electrical conductivity to
make devices like solar cells.

II.1.1 Exercise
Indium tin oxide (ITO) typically has a bandgap of at least 3.5 eV.
1. Please show your calculations − simply listing a number will NOT receive any credit!
(a) What wavelength light is needed to excite ITO’s electrons from the valence to the
conduction band?
(b) In what part of the electromagnetic spectrum does this light fall? (eg. gamma ray,
radio, etc.).
2. Sketch a graph of the solar radiation spectrum versus wavelength, being sure to label your
axes, and label the wavelength at which this graph peaks. Label the boundaries of the
visible spectrum and the wavelength from #1a above on this solar spectrum graph.
(a) In what part of the electromagnetic spectrum does the peak fall?
3. For reasons that will be clearer later, the ideal bandgap for a basic solar cell is 1-1.5 eV,
depending on various factors (such as whether the solar cells will be on Earth or in space).
(a) What is the wavelength of light corresponding to this ideal bandgap?
(b) In what part of the electromagnetic spectrum does this light fall? (eg. gamma ray,
radio, etc.).
(c) Will the ideal solar cell material absorb a meaningful amount of light with a longer
wavelength than what you found in part 3a? Why or why not?
(d) Given that a solar cell needs to absorb photons to generate electricity, what regions
of the electromagnetic spectrum should a solar cell absorb to function?
(e) Will ITO absorb a meaningful amount of light at the ideal solar cell wavelength that
you found in part 3a? Why or why not?
(f) Will ITO absorb a meaningful amount of light at the peak solar spectrum wavelength
that you labeled? Why or why not?

6
 (g) Antireflection coatings (ARCs) are often used to improve solar cell efficiency by
reducing the amount of reflection from the surface and allowing greater absorption
of this incoming sunlight. A good antireflection coating will be transparent to the
wavelengths that we want the solar cell itself to absorb.

Based on this, would ITO make a good ARC? Why or why not?

II.2 p vs. n-type Semiconductors

Modern optoelectronic devices rely on semiconductors with impurities deliberately introduced
into the core material. The resulting impurity-infused semiconductor can be referred to as either
a p-type semiconductor or an n-type semiconductor.

A p-type semiconductor has an impurity with fewer valence electrons than the core material,
while the n-type semiconductor has an impurity with more valence electrons than the core
material. For example, silicon has 4 valence electrons. A p-type silicon layer contains atoms
like boron, which has 3 valence electrons, while an n-type silicon layer contains atoms like
phosphorous, which has 5 valence electrons (see this periodic table for reference). This means
that if you put a p-type and n-type semiconductor layer next to each other, the n-type side
will have an excess electron concentration while the p-type side will have a reduced electron
concentration. For the purposes of calculations, this missing electron concentration is treated
like an excess concentration of a positively-charged ”particle” called a hole. A hole is not
a physical positively-charged particle that can move, but it can be treated as such for the
purposes of calculating a system’s electrical properties. When an electron from a neighboring
atom moves into the empty spot represented by a hole, it leaves behind its own empty spot,
which is effectively equivalent to a hole having moved from one atom to another. Figure 2 below
illustrates this phenomenon of electron and hole motion:

Figure 2: Illustration of electron and hole motion. In both cases, electron motion is the physical mechanism for particle
motion, but for calculation’s sake, holes are treated like physical particles.

The act of deliberately inserting impurity atoms into a material is called doping, and the
impurity atoms themselves are referred to as dopants to distinguish them from unintentional
impurities.

Physically, an n-type dopant has valence electrons at a higher energy compared to the semi-
conductor’s valence electrons, while a p-type dopant has holes at a lower energy compared to
the semiconductor’s conduction band (which remember, is relatively empty for a semiconductor
at room temperature). This makes it MUCH easier for electrons from an n-type dopant to be
excited into the semiconductor’s conduction band and for holes from a p-type dopant to be
excited into the semiconductor’s valence band, so much so that for calculation simplicity we
can effectively assume this to occur with 100% efficiency. Below is a diagram representing the
physical mechanism behind doping:

7
Figure 3: Illustration of dopant energy levels in n-type and p-type semiconductors (extra electrons and extra holes,
respectively).

The Fermi level will be discussed later in this section, so you can ignore that for now. For a
Refer to Section 1.2 of Hu’s book for more detail.

II.2.1 Exercise
Consider indium phosphide (InP), a compound semiconductor.
1. To what column of the periodic table does indium (In) belong? To what column of the
periodic table does phosphorous (P) belong?
2. From what column(s) of the periodic table would you want to choose a doping element to
create p-type InP? From what column(s) would you want to choose a doping element to
create n-type InP?

II.3 Density of States

All you need to know for this idea is that quantum mechanics dictates the states a particle or
system can be in. For electrons in a semiconductor, the density of states is zero in between
the conduction and valence bands, while it is nonzero above the conduction bands and below
the valence bands. The density of states is useful in calculating carrier concentrations within
semiconductors. You can find more quantitative information in Section 1.6 of Hu’s book if you
would like. A derivation of the density of states requires knowledge of quantum mechanics, but
can be found here for interested readers.

II.3.1 Exercise
As you can see in the references in this section,
Dc (E) ∝ mn3/2 and Dv (E) ∝ mp3/2
where mn is the effective electron mass and mp is the effective hole mass. These are
quantities that simplify the effects of local electron and ion interactions on carrier motion
equations by pretending that that the electron or hole is an object with mass mn and mp ,
respectively. The exact values of mn and mp vary from material to material, and most are
measured experimentally.
1. Consider silicon, which has mp = 0.59me at a temperature of 4 K and mp = 1.15me at
a temperature of 300 K, and an electron effective mass of mn = 1.06me (me is the rest
mass of an electron in free space).
(a) At a given energy E below the valence band energy Ev , at which temperature will
the density of states be higher? Justify your answer.
(b) At a temperature of 4 K, will the conduction band or valence band have a higher
density of states? Justify your answer.

8
II.4 Fermi Levels
Thoroughly understanding this section from first principles requires an understanding of sta-
tistical thermodynamics. This section will take the Fermi-Dirac distribution as a given for
electrons.
The Fermi-Dirac distribution function describes the probability that a fermion (particle
with half-integer spin) occupies a state with a particular energy. An electron, with a spin of
±1/2, is a fermion (as would a particle with spin ± 3/2, etc., but not a particle with spin 0,
±1, etc.). It is given by the following function:

1
f (E) = (1)
1+ e(E−EF )/(kT )
where E is the energy of the fermion, typically defined with respect to the top of the valence
band (so E = 0 at the top of the valence band). EF is dictated by the total number of particles
in the system. Notice that this function does not depend on the properties of the material − it
applies for any system of non-interacting, indistinguishable particles with half-integer spin (like
non-interacting electrons). Notice also that for an electron with energy E = EF ,

1 1
f (E) = f (EF ) = 0
=
1+e 2
Thus, for any system described by the Fermi-Dirac distribution, the probability of an electron
having energy E = EF is 1/2 (there’s discontinuity issues at T = 0, but we will not concern
ourselves with this for the sake of the exam).
In the conduction band, we are concerned with the probability of an electron occupying a
particular state, which is given by fC (E) (Fermi-Dirac distribution in the conduction band).
In the valence band, we care about holes for calculation purposes, and therefore are concerned
with 1 − fV (E), the probability that an electron DOES NOT occupy a particular state in the
valence band.
Because the Fermi energy is dictated by the total number of electrons and holes, the Fermi
energy of a system in equilibrium (no applied voltage, no light shining on it, etc.) is constant
throughout a system. When a semiconductor is doped p-type or n-type, this shifts the Fermi
level because the dopant impurities increase the probability that an electron will occupy the
conduction band or that a hole will occupy the valence band.
Interested readers can refer to this forum for an explanation of the Boltzmann distribution, a
building block of statistical thermodynamics, and this derivation of the Fermi-Dirac distribution.

II.4.1 Exercise
1. Consider a state in the conduction band that is 0.6 eV above the Fermi level EF .
(a) What is the probability that an electron at room temperature T = 300 K will occupy
this state? What about at T = 0.1 K? Please show your work to get credit.
2. Now consider a state in the valence band that is 0.6 eV below the Fermi level EF .
(a) What is the probability that an electron (NOT a hole!) at room temperature
T = 300 K will occupy this state? What about at T = 0.1 K? Please show your work
to get credit.
3. Based on the previous two questions, sketch graphs of the occupation probabilities f (E)
as a function of the state energy E at both T = 300 K and T = 0 K (yes, the T = 0 K
case is a limiting case for a physical system, even though mathematically you can’t just
plug in T = 0 K to figure this out). Label f (EF ) on both graphs.

9
II.5 Electron and Hole Concentrations in Semiconductors
Check out Section 1.8 of Hu’s book for a derivation of electron concentrations n and hole
concentrations p in semiconductors as a function of energy E above or below the valence band. In
summary, the concentration of electrons (electrons per unit volume) occupying a state between
energy E and E + dE is given by
D(E) ∗ f (E) dE
since there are D(E) dE state between E and E + dE. You need both the density of states and
the probability of a carrier occupying said state. If there are no available and allowed states at
a particular energy E, then the electron cannot occupy them and actually reside at that energy
even if f (E) ̸= 0. The relevant formulas for carrier concentration to remember from Section 1.8
of Hu’s book are
2πmn kT 3/2
 
−(Ec −EF )/(kT )
n = Nc e where Nc ≡ 2 (2)
h2
for electron concentration in the entire conduction band (all energies E ≥ Ec ) and
 3/2
2πmp kT
p = Nv e−(EF −Ev )/(kT ) where Nv ≡ 2 (3)
h2

for hole concentration in the entire valence band (all energies E ≤ Ev ).

The product np in a semiconductor at equilibrium (no potential bias or illumination) can be
written as
np = n2i (4)
where ni represents the concentration of the electrons in the conduction band and holes in the
valence band for an intrinsic semiconductor in equilibrium (no potential bias or illumination).
For an intrinsic semiconductor, n = p = ni because for a positive temperature T , electrons and
holes form in pairs when an electron is excited into the conduction band and leaves behind a
hole. Note, however, that np = n2i even for doped samples! This is because the excess carriers
from doping the sample (eg. excess electrons in an n-doped semiconductor) recombine with
some of the intrinsic holes, and the recombination rate is proportional to np.

II.5.1 Exercise
1. Based on the above (including Section 1.8 of Hu’s book), what is the electron concentration
at the Fermi energy EF of a semiconductor in which EF is between EV and EC (true
for almost all semiconductors in use)? Justify your answer.
2. Consider a sample of gallium arsenide (GaAs), with effective electron mass mn = 0.063me
(where remember, me is the rest mass of an electron in vacuum).
(a) Calculate D(E) and f (E) at E = 1.47 eV where EC = 1.42 eV and EF = 0.71 eV.
(b) Calculate the electron concentration n in the conduction band of GaAs using Equa-
tion 1.8.5 and Table 1-4 in the book.
(c) When a semiconductor is doped p-type or n-type, the concentration of impurity
atoms added (donors or acceptors) is typically at least 1016 / cm3 .
How does this compare to the electron concentration calculated in 2b? Do you expect
that doping would enhance conduction in a sample relative to an undoped sample?
Think about the orders of magnitudes of the values you calculated in Question 2. Does
it make sense that the electron concentration in the conduction band, n, would be what
you calculated? (Not graded, just for your own understanding).

10
 3. The density of GaAs is 5.32 g/cm3 . Its molar mass is 144.645 g/mol (1 mol = 6.022 ∗ 1023
atoms).
(a) Calculate the atom density per cubic centimeter for GaAs.
(b) How does this compare to the electron concentration you calculated in 2b?
Do you think you would observe significant conduction in GaAs under these conditions?
(Not graded, just for self-reflection).
4. Examine Equations 2 and 3.
(a) Using those equations, calculate the product np. Does np depend on the position of
the Fermi energy EF ? What (other) factor(s) affect np? Note that Nc and Nv are
constant for a given semiconductor at a temperature T .
(b) Calculate ni .
(c) Now consider a sample of n-type InP with donor concentration Nd = 1016 cm−3 .
Undoped InP has an intrinsic carrier concentration of ni = 1.3 ∗ 107 cm−3 . Assuming
complete donor ionization, n = Nd + ni .
Calculate the hole concentration p in this n-type InP sample. How does this compare
to the electron concentration in the n-type InP? How does it compare to the hole
concentration in undoped InP?
(d) Using the information about InP from question 4c, calculate the bandgap of InP
(Ec − Ev ). Show your work.

II.6 p-n Junction

This section will explain the concept of a p-n junction using silicon for simplicity. The con-
cepts translate to more complicated semiconductors as well, but various details, such as those
surrounding the desired impurity types and properties, will be different.
A p-n junction consists of a p-type and n-type semiconductor next to each other. In Section
II, the valence and conduction bands were straight because they were for a pure (crystalline)
material. When impurities are introduced, these valence and conduction bands bend within the
material. Below is a diagram of the band structure of p and n-type semiconductors as well as
a p-n junction:

(a) Band structures of n and p type semiconductors (b) Band structure of a p-n junction

Figure 4: A comparison of the band structures of p and n type semiconductors that are far apart (4a) and next to each
other (4b).

Don’t get too confused by the diagram − there’s a lot going on here that hasn’t yet been
discussed! Additionally, note that the n-type semiconductor is on the left and the p-type on the
right in Figure 4a, but it’s flipped around for Figure 4b. Keep in mind that the band diagram

11
illustrates electron energies, so a higher-energy hole is actually lower in the diagram than a
lower-energy hole because of its opposite charge.
For this example, let’s say the p-type side represents silicon with boron impurity atoms inserted,
while the n-type side is silicon with phosphorous impurity atoms inserted. At equilibrium (no
applied voltage, no sunlight shining on the sample, etc.), the Fermi level needs to be is constant
throughout the sample. This is because once equilibrium is achieved, the number of free carriers
in question remains the same, and the derivation of the Fermi-Dirac distribution specifies that
the position of EF relative to the valence band is dictated by the total number of carriers.
To achieve this, the conduction and valence bands must ”bend” when a semiconductor has a
p-type and n-type region adjacent to each other. Achieving this requires a flow of electrons
from the n-type to p-type region and a flow of holes from the p-type to the n-type region until
equilibrium is achieved.
In this initial stage, the diffusion of majority carriers (electrons from n-type, holes from p-
type) dominates. However, as this diffusion occurs, electrons leave behind positively-charged
ions while holes ”leave behind” negatively-charged ions. See Figure 4-6 in Hu’s book for a
diagram illustrating this charge buildup in the p-n junction. Neither of the left-behind ions are
able to move through the sample, so this creates a build-up of positive charge in the n-type
region and a build-up of negative charge in the p-type region. This results in an electric field
that opposes the diffusion of the majority carriers, since the negative charge in the p-type region
will attract the holes back whlie the positive charge in the n-type region will attract the electrons
back. This results in a drift current that opposes the diffusion current. At equilibrium, the
drift and diffusion currents balance each other out.
The width of this region of left-behind charges, called the space-charge region, is given by
s  
2εs ϕbi 1 1
Wdep = + (5)
q Na Nd

Note that for Na ≫ Nd , Wdep → |xN | and for Nd ≫ Na , Wdep → |xP |.

The amount of band-bending in a p-n junction under equilibrium is given by the built-in
potential:  
kT Nd Na
ϕbi = ln (6)
q n2i
where q is the electron charge and Nd and Na are the donor and acceptor concentrations,
respectively. See Chapter 4 of Hu’s book for a derivation of this expression.
Under a potential bias or illumination, however, the system is out of equilibrium, and the
Fermi energy EF is no longer constant within the sample. See Section 2.8 of Hu’s book for a
more thorough and quantitative treatment of this Fermi level splitting, with concepts such as
quasi-equilibrium and quasi-Fermi levels. From that section, we can express the electron
concentration in the entire conduction band, and the hole concentration in the entire valence
band, using these new quasi-Fermi levels as

n = Nc e−(Ec −EF n )/(kT ) (7)

and
p = Nv e−(EF p −Ev )/(kT ) (8)
where EF n is the electron quasi-Fermi level and EF p is the hole quasi-Fermi level.
For now, I will focus on a qualitative explanation of the physical mechanisms governing free
carrier behavior in these various deviations from an unbiased p-n junction in the dark.
Below is a diagram illustrating the p-n junctions under forward, zero, and reverse bias:

12
 Figure 5: Schematic of a p-n junction under forward bias (A), no bias (B), and reverse bias (C).

The p-type side is on the left and the n-type side is on the right. In (A), the p-type side is
positively biased while the n-type side is negatively biased. In (C), the p-type is negatively
biased while the n-type side is positively biased.

The red circles represent the electron distribution in the conduction band and the white circles
represent the hole distribution in the valence band (numbers not to scale − they simply serve
an illustrative purpose). In (B), when the drift and diffusion currents balance, there’s a certain
number of electrons from the n-type side with energy above Ec in the valence band − it is ONLY
these electrons that can diffuse to the p-type side, since they are above the energy barrier formed
by the band-bending in the conduction band. The red circles with energy less than the Ec on
the p-type side cannot make it over this barrier, and therefore cannot diffuse to the p-type side.
A similar logic applies for the holes in the valence band which are or are not able to diffuse to
the n-type side.

Under a forward bias, this potential barrier, this difference between the conduction band energies
is reduced since the positive bias on the p-type side drags the bands down and the negative
bias on the n-type side pushes the bands up. Thus, a greater fraction of the carriers now
have sufficient energy to overcome the lower energy barrier between the two halves of the p-
n junction, increasing the diffusion current. Meanwhile, the drift current (effectively) doesn’t
change (relative to the diffusion current) because drift current is linearly proportional to electric
field, which is linearly proportional to carrier concentration, while the carrier distribution above
the conduction band and below the valence band is exponential. Thus, there is a greater net
current flow, with electrons flowing to the p-type side and holes flowing to the n-type side.

Under a reverse bias, on the other hand, the negative bias on the p-type side pushes the bands
up while the positive bias on the n-type side pushes the bands down, which increases the energy
barrier between the bands. Now, fewer carriers are able to diffuse over this energy barrier.
The greater potential drop between the p and n-type sides of the junction, as evidenced by
the steeper slope between these regions, does slightly increase the drift current in the ”reverse”
direction.

Finally, under illumination (which is relevant for a solar cell), the band diagram (once the
illuminated system is in equilibrium) looks like this:

13
 Figure 6: p-n junction under illumination

Illumination excites electrons from the valence band into the conduction band, which generates
both excess electrons and excess holes (in pairs) compared to the number of free carriers before.
This increase in the number of free carriers causes Fermi level to split (since the old Fermi level
was based on the previous number of free carriers). Because of the generation of new electron-
hole pairs, rather than just a movement of carriers (like in the biased case discussed previously)
the electrons and holes have their own Fermi levels that reflect the new concentration of each
in the conduction and valence bands, respectively.
Once electron-hole pairs form by absorbing this energy (in this case, from the incoming light’s
photons), they can also recombine. This happens when the electron falls back down from the
conduction band into the valence band and occupies a hole again. This makes the free electron
and hole ”disappear” (since the electron that returned to the valence band is no longer ”free”,
and the hole previously there has now been occupied). This occurs because the electron is then
in a lower-energy state. However, the continued illumination will keep exciting electrons into
the conduction band when it absorbs a photon, so eventually there is an equilibrium between
electron-hole pair generation and recombination. Check out Section 2.6 in Hu’s book for more
information on carrier recombination.
See Figure 4-6 in Hu’s book for diagrams relating to the rest of this paragraph. The depletion
region already has fewer dopant carriers (fewer donor electrons in the n-type half of the depletion
region, fewer donor holes in the p-type half of the depletion region), so electron-hole pairs won’t
really form IN the depletion region. However, near the boundary between the depletion and
neutral regions, electron-hole pairs can be excited by incoming photons. The electrons in the
conduction band of the p-type region will be swept into the conduction band of the n-type
region because of the electric field that causes the conduction band to slope down from the
p-type region through the depletion region into the n-type region. Similarly, the holes that
form at the edge of the n-type region will be swept up into the valence band of the p-type
region because of the potential difference illustrated by the sloping valence band through the
depletion region (remember, for holes, higher up the energy curve is ”lower energy” because of
their opposite charge from electrons). This means that under illumination, there is a non-zero
reverse current when there is no potential bias applied to the system.

II.6.1 Exercise
1. Show that when the p-n junction IS in equilibrium (i.e. np = n2i ), that EF n = EF p using
Equations 7 and 8

14
 2. How does increasing the donor concentration affect ϕbi ? How does increasing the acceptor
concentration affect ϕbi ?
3. Based on your answer to question 2 and the dependence of Wdep on dopant concentration
(and / or Section 4.2.1 from Hu’s book), how does increasing the doping concentration
affect the electric field in the depletion region?
4. The drift current referred to in this section is proportional to the electric field:
dV
Jdrift ∝ E ∝
dx
where V is the electric potential. Based on this,
(a) How would increasing the dopant concentration affect the drift current?
(b) How would this affect the current of a p-n junction under illumination?
(c) Based on this information alone (and ignoring other effects of doping), is a higher or
lower p-n junction doping more desirable for a well-functioning solar cell (one that
generates as much electricity as possible from incoming light)? Justify your answer.

II.7 Current-Voltage Behavior

Please refer to Chapter 4 in Hu’s book for information on the derivations relevant to the formulas
that will be introduced in this section.
For a p-n junction in the dark under bias, the net current (density) flow can be quantified as a
function of applied voltage with the equation

J(V ) = J0 (eqV /(kT ) − 1) (9)

where J0 is a parameter that quantifies the current density under infinite reverse bias, in which
regime drift current dominates. This is also called the saturation current (density). NOTE:
the book will write these formulas in terms of current rather than current density, and I may
occasionally use these terms interchangeably (though I’ll try to avoid it) − the only difference
is that using current density divides out area − I don’t need to repeatedly read about the
dependence of device current on its area. That’s effectively meaningless − it’s like saying bigger
galaxies have more mass.
A positive current (density) reflects ”forward” current flow, in which diffusion current is greater
than drift current, while negative current (density) reflects ”reverse” current flow, in which drift
current is greater than diffusion current (in other words, the direction of diffusion current, the
current flow from the n-type to the p-type region, dictates ”positive” current flow). Notice that
the diffusion current term, which is the first term, is exponential while the drift current term,
which is the second term, is effectively constant as a function of voltage.
For a p-n junction under illumination, the net current (density0 flow can be quantified as a
function of voltage as
J(V ) = J0 (eqV /(kT ) − 1) − Jsc (10)
which almost resembles the equation of a p-n junction in the dark except for the Jsc term, which
reflects the additional drift current that results from the extra electron-hole pairs that form
under illumination and are swept up by the electric field of the depletion region. The value of Jsc
depends on the photon flux density (# of incoming photons per unit area at a particular photon
energy ℏν) and the bandgap of the semiconductor, assuming perfect, complete absorption and
harnessing of every incoming photon at or above the band-gap energy. To quantify this, we can
write ˆ ∞
E2
Jsc (Eg ) ∝ E
dE (11)
Eg e − 1

15
where E is the photon energy. This expression arises because the integrand is proportional to
the solar spectrum (assuming the sun is an ideal blackbody).

II.7.1 Exercise
1. Using Sections 4.9, 2.5, and 2.2 of Hu’s book,
(a) What does J0 depend on? (NOTE: one of the variables you’ll find in Section 4.9
that affects J0 depends on other parameters that are described in Section 2.5, which
then depend on factors mentioned in Section 2.2). Explain thoroughly what each of
these parameters is where needed.
(b) At a given, fixed temperature (eg. room temperature), which of these variables can
be manipulated / improved to achieve a higher J0 ? For each of those, describe
whether you would want to increase or decrease its value and how you can achieve
this.
2. Along with Jsc and J0 , another parameter that tells us about solar cell performance is
called open-circuit voltage (Voc ). This is the value of V at which the current density
through the solar cell is 0:

J(Voc ) = J0 (eqVoc /(kT ) − 1) − Jsc = 0

(a) Derive an expression for Voc in terms of J0 and Jsc . First, don’t make any approxi-
mations, and then make the approximation that Jsc ≫ Js .
(b) A higher Voc usually points to a better solar cell efficiency, all else equal.
Is a higher or lower short-circuit current (density) Jsc desired for a better Voc ? Is a
higher or lower saturation current (density) J0 desired for a better Voc ? Justify your
answers.
(c) Do you think a smaller or larger semiconductor bandgap Eg would improve the Jsc
in the desired direction (based on your answer for the previous part) for higher Voc ?
Justify your answer.
(d) Do you think a smaller or larger semiconductor bandgap Eg would improve the
saturation current J0 ? Justify your answer. (HINT: Section 4.9 of Hu’s book
contains the expression for I0 , and remember the saturation current I0 is the same for
all potential biases and illumination conditions, including a dark system at V = 0).
(e) Using Equation 11 and the expression you found in part 2d,
i. Calculate Jsc (Eg ) and plot it as a function of Eg using Wolfram Alpha (eg. from
Eg = 0 to Eg = 5 in these units).
ii. Then, sketch a plot of J0 vs. Eg (HINT: set all of the constants equal to 1 to
simplify the expression, since we are concerned with the dependence of J0 on
Eg )
iii. Now, sketch a plot of Voc (Eg ) as a function of Eg using the previous two parts.
All in all, does Voc depend on band gap Eg ? If so, how does increasing or
decreasing the band gap affect the Voc ?
(f) Is there a trade-off inherent in optimizing the bandgap of a semiconductorfor maxi-
mizing solar cell efficiency? (A proxy for this is maximizing both Jsc and Voc simul-
taneously). Justify your answer. (NOTE: You must get your answers to both parts
2c and 2d to get credit for this part.

16
 3. The power (energy per unit time) generated by a solar cell is given by

P (V ) = I(V ) ∗ V

Note that since I(V ) < 0 for V < Voc , which means the power is ”negative” − this sign
convention indicates that power is being generated, rather than consumed, in this voltage
range.
Derive an expression for Vmp , the voltage at which the generated power is maximized,
as a function of Voc . (HINT: calculate the derivative of P (V ) and set it equal to 0:
dP/dV = 0, and then use the approximation used in the second part of 2a). Show your
work.

III Organic Semiconductors

Organic semiconductors and devices were developed more recently than their inorganic coun-
terparts, and record organic solar cell performance still lags behind record inorganic solar cell
performance [1]. Despite their lower efficiencies, organic solar cells have several features that
make them more attractive than their inorganic counterparts: they are cheaper, flexible, and
lightweight [2]. Their flexibility and low weight make them easy to integrate into existing in-
frastructure [2]. These advantages are possible because organic semiconductors have absorption
coefficients 1000 times those of polycrystalline silicon and have lower densities, which allows
thin, light layers of active materials to absorb photons efficiently.

III.1 Physics of Electron Arrangement in Atoms

Organic semiconductors have a somewhat different energy structure than inorganic semicon-
ductors. The formation of electronic bands, which allow charge transfer, in organic molecules
can be understood by orbital hybridization theory. Electrons that orbit atoms come in different
orbitals, each of a different energy level. Below is an image that shows the shapes of the orbitals
that are present in carbon atoms:

Figure 7: Schematic representation of electron orbitals in a carbon atom [3]. The arrows represent the arrangement of
electrons and their spins in different orbitals.

The 1s and 2s orbitals, which are represented by the pink sphere in Figure 7 above, are filled
first [3]. Each has two electrons, one of each spin. Then, the higher-momentum 2p orbitals
are filled, but this time, each orientation (px , py , pz ) is filled with one spin-up electron before
being filled with any spin-down electrons. This is due to a combination of the Pauli exclusion
principle and Hund’s rule of maximum multiplicity [4].

Pauli Exclusion Principle

The Pauli exclusion principle states that each electron must have a unique quantum description
− this means that each electron must be in its own energy level and orbital (so either 1s, 2s,

17
2px , 2py , or 2pz ), and that two electrons in the same orbital must have different spins (up or
down). This results because electrons are part of a class of indistinguishable particles called
fermions, for which multi-particle wavefunctions change in sign when particles are renamed [5].
In other words, a two-electron system’s wavefunction ψ is represented as such:


ψ n1 ,n2 ≡ c n1 n2 − n2 n1 (12)

where c is a proportionality constant that helps normalize the wavefunction [5]. Note that the
first represents the ”first particle” and the second represents the ”second particle”, while
the n1 and n2 refer to the state the particle is in. The indistinguishability of the particles
makes the total wavefunction a superposition of the states n1 n2 , in which the first electron
is in the n1 state and the second electron is in the n2 state, and n2 n1 , in which the first
electron is in the n2 state and the second electron is in the n1 state [5]. The wavefunction of a
fermion is antisymmetric because
   
ψ n2 ,n1 = c n2 n1 − n1 n2 = −c n1 n2 − n2 n1 = − n1 , n2

when I switched the order of the state labels n1 and n2 in the wavefunction. The wavefunction
for a distinguishable pair of particles, in which you can tell the difference between the first
and second particle, is simply given by

ψ n2 ,n1
= n1 n2 (13)

when the first particle is in state n1 and the second particle is in state n2 . In general, n1 n2 ̸=
n2 n1 , so Equation 13 does not work for fermions.
The particles obeying the equation


ψ n1 ,n2
≡ c n1 n2 + n2 n1 (14)

for indistinguishable particles are called bosons. Bosons have a symmetric wavefunction:
   
ψ n2 ,n1 = c n2 n1 + n1 n2 = c n1 n2 + n2 n1 = n1 , n2

A quick aside: a wavefunction, at the end of the day, exists to predict the probability of obtaining
various measurements (in quantum mechanics, the values of observables like particle energy are
technically probabilistic). The result is that the square of the magnitude of its integral across all
of the relevant independent variables dictating its behavior (eg. position, time) should evaluate
to 1. In the notation used here, this can be written as

n1 n1 = 1 and n1 , n2 n1 , n2 = 1 (15)

where ... ... represents a multiplication of a wavefunction by its complex conjugate (called an
inner product). For distinguishable particles, for example, we would say that

n1 , n2 n1 , n2 = n1 n2 n1 n2 = n1 n1 n2 n2 = 1 ∗ 1 = 1

You can read more about this in pages like this Wikipedia link on bra-ket notation (used to
write wavefunctions like ), paying special attention to the Hermitian conjugation section
of that page. Additionally, this plus this link on complex conjugates will help (and your own
Googling), but
√ the important takeaway is Equation 15. If n1 = a+bi (where i is the imaginary
number i = −1), then n1 n1 = (a − bi)(a + bi). I used this reference to clarify convention
regarding label order and tensor product calculation. PAY CAREFUL ATTENTION TO
THIS ORDERING TO MAKE SURE YOU EVALUATE THESE EXPRESSIONS
CORRECTLY IN THE EXERCISES!

18
III.1.1 Exercise

Consider a pair of electrons.

1. What happens if you try to put both electrons in the same state (eg. n1 )? Evaluate
Equation 12 in that case.

2. Could you possibly properly normalize this result? (i.e. can you use the result of the
previous question to obtain Equation 15)? Show your work and / or justify your answer.

3. Based on this, do bosons follow the Pauli Exclusion Principle? Justify your answer using
the previous two questions.

III.1.2 Exercise

Now consider a pair of bosons.

1. What happens if you try to put both bosons in the same state (eg. n1 )? Evaluate Equation
14 in that case.

2. Could you possibly properly normalize this result? (i.e. can you use the result of the
previous question to obtain Equation 15)? Show your work and / or justify your answer.

3. Is this consistent with the Pauli Exclusion Principle for fermions? Justify your answer
using the previous two questions.

III.1.3 Exercise

Determine an expression for c (in terms of inner products of n1 n2 ) for

1. Fermions (starting with definition 12)

2. Bosons (starting with definition 14)

Read the following for more background on wavefunctions and symmetries and how they dictate
electron arrangement in atoms such as hydrogen, helium, and lithium. Important takeaways
include:

• To put two electrons in a single orbital (eg. 1s, 2px , etc.), their spins need to be opposite
to construct an antisymmetric wavefunction

• If more than two electrons are in a given orbital (eg. 1s, 2px , etc.), it is impossible to
construct an antisymmetric wavefunction

Hund’s Rule

Meanwhile, Hund’s rule states that electrons are arranged in orbitals to maximize spin [4].
Because electrostatic repulsion causes two electrons in the same orbital to have a higher energy
than each electron in a different orbital, and the atom wants to minimize its energy, each orbital
in a given energy level is filled with one electron before any orbital in that energy level is filled
with another [4]. In this electron arrangement within a carbon atom, two p orbitals are partially
filled, and one is empty.

Read the section Relative stability of 2p2 configurations in the linked PDF for a more detailed
explanation below, but I’ll summarize below.

19
There are five possible wavefunctions to represent the spin configurations in a carbon atom [5]:
h i
ψ↑x ↓x = 2px (1)2px (2) × ↑ (1) ↓ (2)− ↑ (2) ↓ (1)
h i h i
ψ↑x ↓y = 2px (1)2py (2) + 2px (2)2py (1) × ↑ (1) ↓ (2)− ↑ (2) ↓ (1)
h i
ψ↑x ↑y = 2px (1)2py (2) − 2px (2)2py (1) × ↑ (1) ↑ (2)
h i h i
ψ↑x ↑y = 2px (1)2py (2) − 2px (2)2py (1) × ↑ (1) ↓ (2)+ ↑ (2) ↓ (1)
h i
ψ↑x ↑y = 2px (1)2py (2) − 2px (2)2py (1) × ↓ (1) ↓ (2)

All of the wavefunctions are formed as products of a symmetric and anti-symmetric wavefunction
[5]. When the spatial wavefunction is symmetric, the spin wavefunction is antisymmetric and
vice versa. Because of the arguments made in the PDF regarding electrostatic repulsion in
these different cases, combined with the magnetic interaction being much weaker and thus not
influencing the outcome, the last three wavefunctions ψ↑x↑y end up being the most stable.

III.2 Molecular Bonding

Just as with atomic orbitals, we can write Schrodinger’s equations for electrons in molecules.
Solutions to these equations can be approximated by writing wavefunctions that are linear
combinations of atomic orbitals. For a diatomic molecule, this gives us

Ψ = ca ψa + cb ψb (16)

where Ψ gives the molecular wavefunction, a and b denote the different atoms, and ca and cb
are proportionality constants that affect the wavefunction normalization [4]. Molecular orbitals
form when atomic orbitals overlap, yielding a relatively high probability that both atoms’
electrons will be found in the overlap region between the nuclei [4]. Bonding requires three
factors: orbitals must be symmetric such that regions with the same sign of ψ must overlap,
atomic orbital energies must be similar so that the formation of molecular orbitals changes the
electron energy to make bonding favorable, and the distance between the atoms must be short
enough for good atomic orbital overlap but long enough that electron repulsion forces do not
interfere [4]. If one of these three conditions is not true, then nonbonding orbitals tend to form
when the atoms come together [4].
When two s orbitals overlap, as in a diatomic molecule like H2 , then the resulting larger molec-
ular electron clouds are linear combinations of the atomic orbitals:
h i
Ψ(σ) = N cd ψ(1sa ) + cb ψ(1sb )
h i
Ψ(σ ∗ ) = N cd ψ(1sa ) − cb ψ(1sb )

where N is a normalizing factor for Ψ while σ and σ ∗ indicate orbitals that are rotationally
symmetric about a line passing through both nuclei [4]. The wavefunctions show that combining
the atomic orbitals yields one lower-energy bonding molecular orbital, with an increased electron
concentration between the nuclei and one higher-energy antibonding orbital, which yields a
node with zero electron density between the nuclei [4]. In bonding orbitals, the electrons cluster
between the nuclei and attract the nuclei towards themselves, which holds the molecule together
[4].
Each p orbital, however, has separate regions whose wavefunctions have opposite signs. Once
again, when two regions that overlap have the same sign, then the electron probability in the
overlapping region is increased, but when two regions that overlap have opposite signs, then

20
the overlapping region has a reduced electron probability [4]. However, because of the shape
and multiple orientations of the p-orbitals, we end up with not only σ and σ ∗ molecular orbitals
but also π and π ∗ molecular orbitals [4]. Below are diagrams which demonstrate the bond
formation:

(a) Schematic of σ bonds (b) Schematic of π bonds

Figure 8: Schematic diagram demonstrating σ and π bonding and anti-bonding orbitals [4]. The white and green orbitals
represent those having wavefunctions of the opposite sign.

As evident in Figure 8a, adding the pz orbitals of the two atoms causes wavefunctions of the
opposite sign to overlap, which leads to anti-bonding orbitals [4]. Meanwhile, subtracting the
orbitals flips the sign of the second pz orbital such that wavefunctions of the same sign are
overlapping (the colors of pz (b) flip, so there’s a same-colored ”blob” that forms in the middle
that demonstrates increased electron probability between the atoms, and thus ”bonding”)[4].

III.2.1 Exercise

1. Examine Figure 8b. Explain what happens (in terms of overlap of wavefunctions with the
same sign vs. opposite sign), similar to my explanation above for pz , for:

(a) px (a) + px (b) (adding two px orbitals)

(b) px (a) − px (b) (subtracting two orbitals)

2. Now think about the implications for bonding and antibonding:

(a) Bonding results for which one(s) of the two?

(b) Antibonding results for which one(s) of the two?

When orbitals of the same type overlap, then the result is a symmetric arrangement of orbitals
such as below:

21
 Figure 9: Schematic of molecular orbitals for atoms with atomic number up to 10 [4].

In the 2p orbitals of Figure 9 above, π molecular orbitals have higher energies than σ while π ∗
orbitals have a lower energy than σ ∗ [4]. Thus, the gap between the π orbitals is lower than
that between σ orbitals.

Non-identical orbitals can mix if their energies are similar enough and they exhibit the same
symmetries. For instance, the σg orbitals of the 2s and 2p energy levels can mix, as can the σu∗
orbitals. Orbital mixing lowers the energy of the lower-energy orbital and raises the energy of
the higher-energy orbital. Below is a diagram illustrating the effects of such orbital mixing:

22
Figure 10: Schematic of the effects of σg orbital mixing between the 2s and 2p levels [4]. The σg (2p) energy level is raised
while the σg (2s) energy is lowered [4].

After orbital mixing, the σg (2p) energy level is now higher than the πu (2p). The exact orbital
overlap dictates which of the orbitals is the highest occupied molecular orbital (HOMO), the
highest-energy bonding orbital in a molecule mostly full of electrons, and which is the lowest
unoccupied molecular orbital (LUMO), the lowest-energy anti-bonding orbital containing few
(if any) electrons at room temperature. However, the HOMO will always be denoted by the
highest-energy non-asterisk orbital (in Figure 10, the σg (2p) orbital) while the LUMO will
always be denoted by the lowest-energy asterisk orbital.

There are other molecular orbitals that form with the overlap of these orbitals with the d
molecular orbitals, as well as nonbonding that occurs in the conditions described earlier, but
the formation of σ and π bonds are the most important to understand organic semiconductors.

III.3 Energy Band Formation

Diatomic molecules contain bonding and anti-bonding orbitals within a single region in between
the two atoms. In the case of organic polymers, the orbital formation leads to the presence of
a quasi-continuous region of bonding and anti-bonding orbitals. Below is a schematic of this
phenomenon:

23
Figure 11: Schematic representing the formation of HOMO and LUMO levels as alkenes get longer [6]. The lowest energy
optical transition lies between the top of the HOMO levels and the bottom of the LUMO levels.

As opposed to inorganic semiconductors, which are held together by stronger covalent bonds in
a crystalline structure, organic semiconductors are made of molecules held together by weaker
Van der Vaals forces in chains and sheets. Van der Waals forces include the electron repulsion
due to the Pauli exclusion principle explained earlier, the attractive and repulsive electrostatic
interactions between ions, dipoles, quadrupoles, and other general multipoles [7]. As a result, the
energy splittings actually reflect overlapping molecular orbitals, rather than the strong atomic
orbital overlap present in inorganic semiconductors. The energy bands that form in organic
semiconductors result from the overlap of π orbitals, as opposed to the bands that form due to
the potentials of overlapping periodic lattices. The rest of the information in this subsection is
a technical aside, and you can skip to subsection III.4.

Technical Aside

As the alkene chains get longer, the interactions between atoms and molecues becomes more
complex [6]. Because of the Pauli exclusion principle, as established earlier, no two electrons
can be in the same state [6]. As a result, the orbitals form at different energy levels but
get closer together, becoming more “bunched up” as evident in the difference between the
orbitals in butadiene and octatetrane in Figure 11. Eventually, for long polymer chains, the
bonding orbitals become quasi-continuous, as do the anti-bonding orbitals. Then, the only
significant gap left is that between the HOMO and the LUMO, which is the bandgap in an
organic semiconductor.

Below is a schematic that demonstrates the band formation from overlapping π orbitals in
benzene:

24
 Figure 12: Schematic of band formation in a benzene molecule [3].

This electron delocalization makes sense in the context of molecular orbital theory discussed
earlier. Going back to Figure 8b, the diagram shows the formation of delocalized bonding
regions when px orbitals overlapped, such that regions of like signs combined to form a large
region within which electrons could move. This is in contrast to the combined orbital shape in σ
bond formation, where both σ and σ ∗ orbitals trap the electrons between atoms. Even if many
σ or σ ∗ regions overlapped, electrons would always be trapped between two nuclei at best, but
if many px orbitals overlapped side-by-side as shown in Figure 8b, then the delocalized regions
would be larger and extend the length of the molecule, allowing for band-like transport.

In practice, some highly-crystalline organic semiconductors, like crystallized rubrene, have a

regular enough structure to allow for such band-like transport. However, usually transport
bands in organic semiconductors are not as smooth and straight in energy space as those of
inorganic semiconductors. Thus, instead of a band transport model to describe carrier motion
akin to that used to describe inorganic semiconductors, a “hopping transport” model involving
phonon-assisted tunneling is used.

III.4 Absorption and Emission

For small and medium-sized organic semiconductor molecules, the conduction and valence en-
ergy levels do not really come in bands, but rather disjoint energy levels because the disorder
in the semiconductor material yields uneven, unstructured molecular orbital overlap. Such
molecules can be modeled as anharmonic oscillators, where the asymmetry results from the
leveling off of the molecular energy due to the atoms being sufficiently far apart (accurately
predicts electrons being able to dissociate from and escape their host atoms if they absorb
enough energy, which a harmonic well would not allow for). The energy wells of interest in
these anharmonic oscillators are formed by the energy level (HOMO or LUMO level) in ques-
tion, and the energy levels within these oscillator potential wells are the vibrational levels in
the molecule. Below is an energy schematic illustrating this [3]:

25
Figure 13: Illustration of the anharmonic oscillator model of vibrational energy levels in organic semiconductors. E0 is the
HOMO, E1 is the LUMO in this case.

To properly understand this schematic, you need to understand that the anharmonic potential
wells (the deformed parabolas) are plotted on these energy vs. nuclear coordinates axes, but
the wavefunctions (sinusoid shapes within the potential wells) are NOT plotted on these axes
− the curves reflect deviations of the wavefunction from 0, and both positive and negative
deviations reflect a higher probability of electron being at a particular nuclear coordinate for a
given vibrational level ν or ν ′ (which you can also see with the orange shaded regions).
Because optical transitions are on the timescale of 10−14 s, which is much shorter than the
timescale of nuclear coordinate motion, optical transitions are portrayed as vertical in Figure
13 above. This is the justification for the Franck-Condon principle, which states that optical
transitions are most likely when there is high vibrational wavefunction overlap for the different
energy levels [3]. In this case, ”wavefunction overlap” refers to the integral
ˆ
ψν ′ ψν ′′ = dq ψν∗′ (q)ψν ′′ (q) (17)

which you can visualize partly by picturing the amount of overlap (of the the shaded orange
regions) if the horizontal lines of the desired wavefunctions (eg. ν ′′ = 0 and ν ′ = 1) are laid
on top of each other (in Figure 13, the vertical positions of ν ′′ = 0 and ν ′ = 1 reflect the
energies of these levels, not the actual values of the wavefunctions − it’s the deviation from the
horizontal line that reflects the wavefunction ). This overlap integral calculates the probability
of a transition from state ν ′′ to ν ′ .
The consequence of this is that even photons with energy higher than the bandgap cannot always
be absorbed. The higher the value of the overlap integral Equation 17, the more probable a given
transition. Check out this link on the Franck-Condon principle for a more detailed explanation
of the effects of overlap on the energy spectra.

III.4.1 Exercise
√ √ √ √
1. Graph S00 / π and S01 / 2 from this link (factors of π and 2 for normalization pur-
poses) as a function of bond length change Re − Qe with α = 1 (you can do this by setting

26
 Re − Qe = x and graphing the result, as long as you remember that x is a CHANGE in
bond length, and not a coordinate in absolute space).
2. What is the most likely change in bond length for the ν ′′ = 0 to ν ′ = 0 transition?
(providing the value of x will suffice, and no need for units). Is this reflective of a vibration?
3. What is the most likely change in bond length for the ν ′′ = 0 to ν ′ = 1 transition?
(providing the value of x will suffice, and no need for units). Is this reflective of a vibration?
√ √
4. If you consider the maximum possible values of S00 (x)/ π and S01 (x)/ 2 across all bond
length changes, which transition is more likely?

III.4.2 Exercise
Earlier in the exam, you determined whether the Jsc and Voc of a solar cell would depend on
band gap or not, and if so then how raising or lowering said bandgap would affect each.
1. Based on Figure 13 and this page on the Franck-Condon principle, does a nuclear dis-
placement between the ground and excited state of such a system effectively increase or
decrease the bandgap? Justify your answer.
The Franck-Condon principle explains Kasha’s Rule, which states that emission typically only
happens from the lowest excited state, regardless of the excitation energy [8]. Because there is
high spatial wavefunciton overlap between the higher excited states and the lower excited state,
such as between v ′ = 2 and v ′ = 0, such intra-energy-level transitions happen quite readily. The
overlap is greatest when not only position but momentum overlap, which means that electrons
in higher excited states will relax most readily into the lowest excited state before fluorescing
by relaxing to a vibrational state in a lower energy level. As a result, for the most part the
emission wavelength is independent of the absorption wavelength.

III.4.3 Exercise
One implication of Kasha’s Rule is that carriers that are excited into a higher vibrational state
(eg. from ν ′′ = 0 to ν ′ = 1) will likely relax down to the lowest excited state (i.e. ν ′ = 0) before
they can be collected to generate a solar cell current J(V ) at a given voltage.
1. How would this phenomenon affect solar cell efficiency, and why?
2. Which of the fundamental solar cell losses, as described in this paper, is represented by
this phenomenon? Explain.

IV Excitons
When charge carriers are excited in organic semiconductors, they form excitons, which are
electron-hole pairs bound by the Coulomb force. In this section, we will explore how excitons
impact organic solar-cell efficiency compared to inorganic ones.

IV.1 Exciton Radii

Figure 14 shows diagrams of two different types of excitons − Wannier excitons and Frenkel
excitons. An exciton consists of an electron orbiting a hole, which means to first order, we
can semi-quantitatively treat the electron as if it is in a hydrogen atom. It can be shown that
the most likely distance between the lowest-energy electron and the nucleus, in the quantum-
mechanical description) of a hydrogen atom is

4πϵ0 ℏ2
r0 = a0 where a0 = (18)
µe2

27
a0 is the Bohr radius, the radius for a hydrogen atom that Bohr predicted (accidentally getting
it right despite his quantum-mechanical description being wrong). µ is the reduced mass of the
two-particle system (electron-proton for the hydrogen atom, electron-hole for the exciton). The
Bohr radius is effectively dictated by the interplay between the Coulomb potential energy and
the overall kinetic energy of the electron-proton system, although fully illustrating this would
require quantum mechanics beyond the scope of this exam (but you can check out derivations
of the hydrogen atom equations, energies, etc. for a full description).
The average electron distance, on the other hand, is calculated in quantum mechanics in a
slightly different fashion:
ˆ ∞
∗ 1
rψ100 4πr2 dr where ψ100 = p 3 e−r/a0


r̄0 = ψ100 (19)
0 πa0

where the 4πr2 term comes because we are integrating over spherical coordinates.

IV.1.1 Exercise
1. Calculate the integral in Equation 19 to evaluate r̄0
2. Is this value equal to the most probable electron distance calculated earlier? Does this
make sense? Explain why or why not using the probability distribution sketched in this
reference.

(a) Schematic of a Wannier exciton (b) Schematic of a Frenkel exciton

Figure 14: Side-by-side comparison of Wannier vs. Frenkel excitons [9].

Independent of your answers to the previous exercise, you should note that both r0 ∝ a0 and
r̄0 ∝ a0 (which should make sense from a dimensional analysis perspective). For an exciton, we
have an electron-hole system within a material, rather than an electron-proton in free space.
The electronic interactions with the material can be summed up by a modification of dielectric
constant used to express the Coulomb potential energy in free space. We replace ϵ0 with
ϵ = ϵr ϵ0 where ϵr varies by material and can be measured. Then, the Coulomb potential of a
hole in a semiconductor becomes
e2
V (r) = (20)
4πϵr

IV.1.2 Exercise
Based on the previous description about the radius of a hydrogen atom and the factors impacting
it,
1. How does modifying this Coulomb potential, as done in Equation 20, change the radius
of the exciton? Express the new radius a in terms of a0 .
2. Since the dielectric constant ϵ reflects the extent to which atoms are polarized in the
presence of an electric field, with higher dielectric constants indicating greater polarization,
is it possible for ϵr < 1? Explain.

28
The energy of such an exciton with radius a (analogous to the energy of the ground-state
electron of a hydrogen atom) is given by

ℏ2
E=− (21)
2µa2

where the energy is negative because E = 0 is defined as the energy of an electron an infinite
distance away from a positive charge.
Since the binding energy of an exciton depends on electron-hole distance, it is not tied to the
material band gap per se. In particular, it can be (and often is) less than the band gap energy.

IV.1.3 Exercise
1. Using Equation 21, describe the dependence of the exciton binding energy on the relative
permittivity ϵr .
2. Inorganic semiconductors like GaAs tend to have larger ϵr while organic semiconductors
tend to have much smaller ϵr . Based on this,
(a) Are inorganic semiconductors more likely to have Wannier excitons or Frenkel exci-
tons?
(b) Are organic semiconductors more likely to have Wannier excitons or Frenkel excitons?
(c) What does this say about the strength of the Coulomb interaction between electrons
and holes in inorganic vs. organic semiconductors?
(d) Are exciton binding energies likely to be higher in organic or inorganic semiconduc-
tors?
3. Charge carriers should be able to move freely in order to be efficiently captured by a solar
cell to generate power. Based on the answers to the previous questions about exciton
binding energy, can charge carriers be more effectively captured in inorganic or organic
semiconductors? Justify your response. (HINT: Excitons can absorb additional energy
to excite the electrons and holes enough for them to separate from one another, but this
additional energy absorption entails losses).

V Data Analysis Primer

When discussing the efficiency of solar cells, two different figures are used to better describe the
device and highlight which factors cause the most significant deviations from 100% efficiency.
The internal quantum efficiency (IQE) reflects the fraction of absorbed photons that generate
current, while the external quantum efficiency (EQE) indicates the fraction of incident photons
that give rise to current [10]. This means that a device with good charge transport properties but
low absorption, either because the surface is very reflective or the material is very transparent,
will have a high IQE but a low EQE. In this section, we will be looking at experimental data
of EQE as a function of incoming photon energy to determine which photon energies give rise
to a good amount of device current.

V.1 External Quantum Efficiency

Recent work has demonstrated that the interface between an organic donor of special interest,
crystalline rubrene, and one particular organic acceptor, C60 gives rise to photocurrent (i.e. cur-
rent resulting from the excitation of charge carriers by light) that results from direct excitation
of excitons [11]. Figure 15 shows the EQE at different photon energies in different systems. The
graph shows the abrupt cut-off in EQE at 1.6 eV for C60 , which is its bandgap [11]. Figure

29
 (a) Crystallinity-dictated EQE (b) Amorphous layer-thickness-dictated EQE

Figure 15: EQE as a function of the energy of incoming light. The crystal planar heterojunction has a nonzero EQE for
energies as low as 1.05 eV. 15a shows how an absolute crystalline vs. amorphous device affects EQE, while 15b shows how
varying the thickness of an amorphous rubrene-C60 layer on top of a crystalline layer of rubrene affects the CT state.

15 shows how factors such as using bulk (BHJ) vs. planar heterojunction (PHJ) structures,
crystallization of a structure, and layer thicknesses can affect the EQE curves. Don’t worry
too much about these details, but pay special attention to the difference between the curves of
amorphous vs. crystalline PHJs.

V.1.1 Exercise

Rubrene has a higher bandgap than C60 . Given that EQE measures the ability for electrons
and holes to be excited and collected in the device at a given photon energy,

1. Would you expect a non-zero EQE at energies below 1.6 eV for a rubrene-C60 system?
Why or why not?

2. Based on the answer to the previous part, how does the EQE data provide evidence of
the excitation of exciton excitation?

The arrows in Figure 15b indicate the exciton binding energies, because the EQE starts leveling
off a bit after these arrows. These ”humps”, therefore, represent a sudden jump in the ability of
the device to harness charge carriers because the excitons have been excited, and the electrons
and holes have been freed from each other.

VI Data Analysis

VI.1 Extracting Exciton Energies from EQE

Figure 16 contains plots of all of the EQE data collected:

30
Figure 16: EQE vs. photon energy plot for crystalline and amorphous rubrene devices using Cl6 -BsubPc-Cl and Cl12 -
BsubPc-Cl as the acceptors.

Figure 17 illustrates data with curve fits to more accurately extract exciton energies (I don’t
remember why this data looks different from Figure 16 − bear with me here).

(a) Cl6 -BsubPc-Cl

(b) Cl12 -BsubPc-Cl

Figure 17: Comparison of curve fits for Cl6 -BsubPc-Cl and Cl12 -BsubPc-Cl acceptor molecules.

I used my graduate student advisor’s fitting script to quantify the differences in extracted
exciton energies between systems using amorphous and crystalline rubrene for both the Cl6 and
the Cl12 -based acceptors. Figure 17 display the plots that show the EQE data as well as fitted

31
curves. The curve to which they were fit was
A 2
EQE = √ e−(E+λ−x) /(4πkT ) (22)
x 4πλkT
where x is the incoming laser light energy in eV, E is the exciton energy, and λ reflects the
broadness of the shape. Below is a table that gives the fitted exciton energies obtained by the
scripts:
Amorphous Crystalline Difference
Cl6 -BsubPc-Cl 1.59 eV 1.34 eV 0.25 eV
Cl12 -BsubPc-Cl 1.36 eV 0.98 eV 0.38 eV
C60 1.48 eV 1.1 eV 0.38 eV

The ”amorphous” vs. ”crystalline” columns reflect the use of amorphous vs. crystalline rubrene.
So the top left entry reflects the exciton binding energy in a system containing amorphous
rubrene and the Cl6 -based acceptor.

VI.1.1 Exercise
Based on the data in the table, does the choice of acceptor affect the reduction in exciton energy
observed when going from amorphous to crystalline rubrene? Explain.

References
[1] Wenchao Zhao, Sunsun Li, Huifeng Yao, Shaoqing Zhang, Yun Zhang, Bei Yang, and
Jianhui Hou. Molecular optimization enables over 13% efficiency in organic solar cells.
Journal of the American Chemical Society, 139(21):7148–7151, 2017. PMID: 28513158.
[2] Henning Richter Barry P. Rand. Organic Solar Cells: Fundamentals, Devices, and Upscal-
ing. Pan Stanford Publishing, Reading, Massachusetts, 2014.
[3] Barry Rand. Thin film solar cells: Organic pv.
https://blackboard.princeton.edu/bbcswebdav/pid-2263573-dt-content-
rid-5958814 1/courses/ELE557-ENE557 S2018/Topic%2011%20-
%20Thin%20film%20solar%20cells%20-%20OPV.pdf, February 2018.
[4] Paul J. Fischer Gary L. Miessler and Donald A. Tarr. Inorganic Chemistry. Pearson, Upper
Saddle River, New Jersey, 5th edition, 2014.
[5] Dan Dill. Many-electron atoms: Fermi holes and fermi heaps.
http://quantum.bu.edu/notes/QuantumMechanics/FermiHolesAndHeaps.pdf, November
2004.
[6] Cambridge Cavendish Laboratory. Organic Semiconductors.
https://www.oe.phy.cam.ac.uk/research/materials/osemiconductors.
[7] E. Hadjittofis, S.C. Das, G.G.Z. Zhang, and J.Y.Y. Heng. Chapter 8 - interfacial phenom-
ena. In Yihong Qiu, Yisheng Chen, Geoff G.Z. Zhang, Lawrence Yu, and Rao V. Mantri,
editors, Developing Solid Oral Dosage Forms (Second Edition), pages 225 – 252. Academic
Press, Boston, second edition edition, 2017.
[8] Paul Suppand. Chemistry and Light. The Royal Society of Chemistry, Thomas Graham
House, The Science Park, Cambridge, 1994.
[9] MIT OCW Scholar. Excitons – Types, Energy Transfer.
https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-973-organic-
optoelectronics-spring-2003/lecture-notes/7.pdf, Spring 2003.

32
[10] Joeson Wong, Deep Jariwala, Giulia Tagliabue, Kevin Tat, Artur R. Davoyan, Michelle C.
Sherrott, and Harry A. Atwater. High photovoltaic quantum efficiency in ultrathin van der
waals heterostructures. ACS Nano, 11(7):7230–7240, 2017. PMID: 28590713.
[11] Michael A. Fusella, Alyssa N. Brigeman, Matthew Welborn, Geoffrey E. Purdum, Yixin
Yan, Richard D. Schaller, YunHui L. Lin, Yueh-Lin Loo, Troy Van Voorhis, Noel C.
Giebink, and Barry P. Rand. Band-like charge photogeneration at a crystalline organic
donor/acceptor interface. Advanced Energy Materials, 8(9):1701494, December 2017.

33