---

Building and testing drug delivery systems

Drug delivery is the field that investigates methods to provide local, targeted, and sustained delivery of molecules. The approaches can range from a large implant that slowly delivers a molecule through a membrane to nanoparticles that are designed to be administered systemically (in the blood) and hone to a particular tissue such as a tumor.

Introduction

Some of the classic approaches for delivering a molecule are outlined in the schematic below.

Figure 1: Retina

I recommend using food coloring for the model drug. The goal is to deliver as much drug as possible for as long as possible from as small a device as possible.

The release time will be defined as the length of time measureable amounts of dye are released.

Designing a Drug Delivery System

You’ll need to sit down and put together a list of input requirements as a starting point for designing a drug delivery system. Input requirements are the list of things that need to be addressed. What is the problem? How does the system need to be able to work to address the problem? If you’re designing a drug delivery system for the brain, you don’t want a device that is bigger than your head, do you? So, you need to think about all the stuff that goes into defining the challenge.

Input Requirements

You literally want to make a list of what your device should do (not what it is.) Focus on a specific application. One example might be that the food coloring is a model for an antibiotic in the eye. Get to know the literature about your desired application.

Hypothetical example for antibiotic in the eye: (this would be a tough set of reqs, so develop your own)

1. Delivery over at least 3 months
2. Loading of the drug of at least 30 mg per gram of device
3. Injectable into intravitreal space using a 33 gauge needle
4. Fully removable post delivery
5. Storage at room temp for 3 weeks
6. Release of drug must be linear
7. Release from device to device must be within 5% of each other

What input requirements fit with your application?

Brainstorming

When you brainstorm, you want to come up with a lot of different possible designs. Don’t edit yourself. How many different ways can you deliver the dye? I would recommend looking at the literature for starting points.

You will want to consider the materials and resources you have as you brainstorm. Some of the materials that are readily available on amazon and safe to use at home include:

• Sodium alginate
• Calcium chloride
• Paraffin wax
• Gelatin

Design Decision Matrix

Once you have a series of potential designs, you need to select which one will address the input requirements the best. A decision matrix is a method for figuring this out. There are loads and loads of different kinds of them. The easiest ones to test are those where you list off all of the input requirements and then weight this importance of each. (Put them in the first column.) After that, you rate the designs on how well they address each input requirement.

Input requirement	Design 1	Design 2…
Delivery over at least 3 months 30%
Loading of 30 mg/g device 30%
Removable 10%
Linear delivery 20%

Verification and Validation

This is where you want to see if you designed fulfills the input requirements. Figure out how much drug is coming out and how much is in there in total. Know how much your system weighs and how it can be delivered in the application.

Doing the Release Study

Take the device and place it in the phosphate buffered saline (PBS). PBS is just water and salts that mimicks the solutions in the body reasonably well. You can get buffered salt tablets at the drugstore or online. You could always use tap water, but it is important to check the tab water to determine the pH. pH paper is sometimes available at drug stores, but I find it most often at amazon or one of my favorite chemical supply companies like arbor scientific or carolina scientific.

Once you have salt solution or PBS, you’ll put your device in the solution and soak it. Since you’ll have it on a table to shelf, the temperature should be room temperature. Try to record the temperature at which you do the release. Most drug delivery systems release faster when they are warm. The body is 37 C (assuming you’re not sick.)

Every hour or two, collect the PBS solution and replace it with fresh PBS. Compare the color of the collected PBS to determine how much has come out. Record the time you took the sample. You’ll figure out how much is released below and will use that to plot the curve for the release. In the meantime, store the PBS you removed in a container.

Quantifying Release

Since you’re using food coloring, you can create a visual scale of release. Essentially, you make vials of known quantities of food coloring and compare your release to the vials to match up each release sample you’ll collect. In the lab, we generally use a plate reader to match things up with more accuracy, but it is impressive what the eye can do, and the most important thing is to be honest and clear about how you do the experiment and what the limitations are. That is always more important than using the fanciest equipment to measure things.

Once you have a number of data points, you need to plot a release curve. In the field, most people plot cumulative release. At your first time point, you should have an amount. Plot it. At the second time point, you’ll have a new concentration. You can add the amount to the first to get cumulative release. As you do this, you’ll generate a curve like the following:

As you can see, for the different formulations, the cumulative release has been plotted. This data is from a project where we created microspheres to delivery timolol maleate, a drug used to lower pressure in the eye to treat glaucoma. The PLGA refers to poly(lactic-co-glycolic acid), a degradable polyester that dissolves in water over weeks. The number refers to the specific polymer. 504 means that it is a 50/50 lactic acid to glycolic acid ratio and 4 refers to how long the chains are. PLGA is a copolymer, and the amount of lactic acid to glycolic acid affects how long it takes to dissolve as does the molecular weight.

I wouldn’t recommend using PLGA at home. The polymer is extremely benign, but one needs to use solvents to process the polymer, and most of these are toxic and require a chemical hood. Instead, I’d recommend using some of the materials described above that can be processed at home safely.

One of the things you’ll notice here is that we have error bars. We did replicates of our samples for release. One of the important questions one should always ask with data and devices is how easily are they replicated? You wouldn’t want a drug delivery system to deliver more drug one time than another. That could cause lots of problems.

---

Make your own solar cells

Much of the information for this lab is drawn from two excellent papers in the Journal of Chemical Education (1, 2)

They provided the foundation for the lab as well as some of the background and protocols. I am beyond grateful. – Lavik

1. Anunson PN, Winkler GR, Winkler JR, Parkinson BA, & Schuttlefield Christus JD (2013) Involving students in a collaborative project to help discover inexpensive, stable materials for solar photoelectrolysis. Journal of Chemical Education 90(10):1333-1340.
2. Smith YR, Crone E, & Subramanian V (2013) A Simple Photocell To Demonstrate Solar Energy Using Benign Household Ingredients. Journal of Chemical Education 90(10):1358-1361.

Websites worth visiting to understand more about solar cells:

• http://www.ucsusa.org/clean-energy/renewable-energy/how-solar-panels-work#.WYtBPVGGPLZ
• https://www.scientificamerican.com/article/solar-cells-prove-cleaner-way-to-produce-power/
• http://solararmy.org
• http://www.pveducation.org/pvcdrom/characterisation/electronics
• https://web.stanford.edu/group/mcgehee/publications/MT2007.pdf

Introduction

Richard Smalley, a Nobel Prize winner, coined the term Terrawatt Challenge in 2004 to describe the call to have 14 TW of renewable energy by 2050 (3). Renewable energy comes in all sorts of packages, but one that has drawn substantial interest in the last decade has been solar cells and solar panels.

The traditional solar panels of my youth were silicon-based systems. The pro is that the energy density of these systems is relatively high. The con is that they are insanely expensive and can break relatively easily. A focus on processing and scale has brought the price down, but they continue to be expensive and their cost coupled to their lifetime issues limits their application.

Meanwhile, organic solar cells have gained great interest. In 1991, O’Regan and Graetzel published “High efficiency solar cell based on dye-sensitized colloidal TiO2 films” in Nature (4), and a world was born. Cheap, flexible, disposable… but were they functional? Their energy densities haven’t exactly been tremendous. So, what now?

How do solar cells work?

Solar cells are photovoltaics. They are essentially the reverse of light emitting diodes. In a photovoltaic, a semiconducting material is used. The materials is doped to create a negative side (n-type) and a positive side (p-type). (With silicon that bonds to 4 atoms, a dopant that bonds to 3 like boron—wow that periodic table is useful—creates a hole, a positive spot. A dopant that bonds to 5 atoms like phosphorous has an extra electron.) When the n type and p type are joined—not actually how they are made, but it is easier to think of it this way—the electrons from the n type move to the holes in the p type. This movement creates an electric field in a portion of the device called the depletion region or depletion zone.)

Photons with energy greater than the bandgap of the material cause electrons to move from the valence to the conduction band. If the electron is created in the depletion region, the electric field will drive it and create a current.

One of the major issues with any kind of solar cell is keeping electrons in the conduction band long enough for them to move and create a current. Electrons fall back into the valence band pretty easily. Defects or imperfections in a material foster this behavior which is called recombination. One of the reasons polycrystalline silicon solar cells are fundamentally less efficient than their single crystalline counterparts is that the interfaces between the crystals, or grain boundaries, are sites for recombination.

One would like to capture as much of the energy from the sun as possible, so, often solar cells are designed with a range of bandgaps or, luminescent material is used to step down the high energy photons to energies consistent with the bandgap of the solar material.

If you attach electrodes to the solar cell or array of solar cells, you can collect the energy from the system. Ta dah! Of course, the moment you do this, you have to find a place to use or store the energy. That’s an important story for another project. First, we need to make some solar cells.

Nature’s solar cells

Solar cells are often compared to plants that undergo photosynthesis. In photosynthesis, chlorophyll converts light into sugars or chemical energy. Solar cells convert light into electrical energy (5). It is not surprising that nature has brilliant approaches to extracting the most energy from light. Many of those techniques have been leveraged in the materials community to increase efficiency of solar cells (5). This approach of borrowing inspiration from nature is often referred to an bioinspired materials.

Silicon based solar cells

The earliest solar cells (photovoltaics) were single crystal silicon p-n junction systems (Figure 1). The beauty of single crystal structures is that one doesn’t have many defects which act as places were charges can get quenched. Single crystal systems are extremely efficient, but they are also very expensive. In the 1990s, a 1 foot by 2.5 ft solar panel would cost well over $1000. People were definitely not putting them on their roofs. It is cheaper to make polycrystalline silicon systems, and as we moved into this century, researchers developed methods to make them efficient enough and cheap enough that the trade off was worth it.

Not to be left behind, the rest of the materials world started asking if their favorite materials could be suitable for photovoltaics.

Dye sensitized solar cells (DSSCS)

These solar cells borrow from nature in that they use a molecule that can absorb light just as chlorophyll does, but the device converts the light to a current. The conversion of the absorbed light is one of the most challenging aspects of these types of solar cells.

The molecules used to absorb light tend to involve ring structures which isn’t terribly surprising. We need molecules that absorb well in the visible and, potentially, UV parts of the spectrum. Two of the most common include the ruthenium-based systems including cis-Bis(isothiocyanato)(2,2′-bipyridyl-4,4′-dicarboxylato)(4,4′-di-nonyl-2′-bipyridyl)ruthenium(II).

One can also use anthrocyanin which comes from blackberries and exist in a number of purple-colored fruits and plants.

Conversion is key with these systems. To convert the energy from the absorbed light into electricity one needs a material with the appropriate bandgap. TiO₂ has a bandgap at 3.2 eV which works well for these molecules. We need as many interfaces as possible between the TiO₂ and the dye molecules, so nanocrystalline TiO₂ particles are preferred with their high surface to volume ratio (6).

Light excites an electron in the dye. The electron is transferred to the titania. If we use an electrolyte solution and connect the system with a circuit, the electron reduces triiodide which then oxidizes and releases the electron to the dye molecule (6).

Experimental methods

Commercial solar cells

I always like to compare the DSSCs to commercial solar cells. One can buy a number of solar cells online at amazon or similar places. In my class, we usually test:

1. A flexible transparent cell
2. A single crystalline cell
3. A poly crystalline cell

What is the difference between these three?

The protocols below will help you assemble DSSCS. You’ll want to quantify the following along the way to make the best assessment possible:

1. I-V curves.
2. Weight of the systems
3. Area of the solar panels.
4. Cost. (Estimate from panels available commercially)
5. Reported efficiency from the literature.
6. Estimate efficiency based on the IV curves you obtain. The following website is extremely helpful in doing this: http://www.pveducation.org/pvcdrom/solar-cell-efficiency
7. Lifetimes of these cells. Look in the literature. How long to the last? What are the mechanisms by which they fail?

Dye-based solar cells

This is adapted from the protocols at the solararmy.org which is a truly phenomenal resource for learning about building DSSCs as well as other solar cells.

One of the reasons DSSCs are so cool is that they are pretty easy to make. Most of the materials can be purchased at hardware stores. However, when I teach the class where we make DSSCs, we use a kit from arbor scientific that was created specifically for the projects in solararmy.org. It makes things a bit easier to have all the pieces designed to go together.

In this set of experiments, the key thing is to come up with a hypothesis about which dye will lead to the greatest power generation and then test the hypothesis. One of the important steps in testing a hypothesis is to make sure that the experiment is designed with the proper controls. If your hypothesis is that a blend of coffee extract and blackberry juice will give you the best outcome, you probably want to test them separately as well as together. You also want to think about your light source. It may be that one formulation works well on a sunny day, but another is better on a grey day. You also may want to make a number of replicates of the same thing to see how much they vary sample to sample.

Preparing the TiO2 electrode

1. Take one piece of the FTO glass and use the multimeter to find the conductive side. Set the multimeter to the resistance setting denoted by the symbol ohm (Ω). If using the model included in the kit, the multimeter dial should just be pointing straight down pointing to 200 ohm.
2. Press the points of the two metal probes onto the surface of the glass, careful that the metal points don’t touch. The conductive side will have a reading around 30 ohms. If you don’t get a reading (i.e. still see the overflow value of “1”), flip the glass over and try the other side.
3. Once you have found the conductive side of the glass (the side with a resistance reading), set that side face up on the table. Take a piece of scotch tape and cover approximately 1/8” of the surface of the glass as shown here. The remaining open surface area will be covered with the TiO2 paste. The taped off strip will be blank glass which is necessary for assembly in the end.
4. Using the plastic pipette, drip a couple drops of the TiO2 solution in the center of the exposed glass.
5. Use the smooth side of the pipette (i.e. no engraved numbers or seam) and immediately squeegee the solution down and up once or twice with the side of the pipette. Aim for a thin, even coating of the paste. If the TiO2 does not coat the entire exposed surface, quickly add another drop of TiO2 paste and re-squeegee the whole plate. It should be a slightly transparent white color. Allow the paste to dry, undisturbed, for a minute or two. Once dry, remove the tape from the glass.
6. Transfer the glass to a hotplate or a frying pan and leave the TiO2 film face up. The exact temperature of the hotplate is not important. Simply the hotter the plate, the faster it will be done.  The surfactant and solvent in the paste will evaporate while on the hotplate, leaving behind just the TiO2 nanoparticles. The glass will appear to turn brown or burned, and then white again.
7. When the glass is done (i.e. the slide turned brown and then back to white), turn off the hotplate and let the glass cool down slowly. If the glass is moved too quickly from hot to cold it will crack. Even touching it with the tweezers can sometimes be enough of a temperature shock to cause cracking. A small crack usually won’t cause problems with the effectiveness of the cell, but best to avoid. The plastic tweezers can also melt so wait for the plate to cool.

Extracting dyes from coffee and blueberries

One can extract dye from a range of plants and berries. In my class, we use coffee and Blueberries as well as blackberry extract that we buy on amazon, but these are really just starting points for thinking about the next steps.

1. Place 0.5 grams of blueberries or coffee in 5 ml of water, isopropanol, or mineral spirits.
2. In the lab, we tend to sonicate this, but since that isn’t practical at home (unless you have a sonicating bath to clean jewelry or parts), I would recommend mincing up the fruit or plant and soaking it overnight. You can also explore heating the solution to see if it improves the extraction (makes it darker)
3. In the lab, we centrifuge the solution to separate the solids. At home, a good coffee fiter can do the trick but may remove some dye. Explore filters versus cheese cloth versus pulling solution using the eye dropper from just the top of the solution that has been allowed to rest for several hours to allow much of the solid to settle. (The centrifuge speeds this process up dramatically)
4. Ideally, you would now characterize the concentration of the dye by absorbance using a UV-Vis spectroscopy system. Unless your home is far cooler than most, it is likely that you don’t have one of these. However, in this day and age, one can set up a simple system using a smartphone. (Yay, technology.) Excellent information on this can be found at: https://www.chemedx.org/blog/use-your-smartphone-absorption-spectrophotometer

To learn more about extraction of dyes, it is worth looking at the following paper: “Potential of Purple Cabbage, Coffee, Blueberry and Turmeric as Nature Based Dyes for Dye Sensitized Solar Cell (DSSC)” (7). This paper is open access, so you should be able to read it for free.

Dyeing the TiO2 electrode and assembling the DSSC

You’ll do this for each of the dyes. Test three different ones per dye solution, if possible, to see how much they vary from device to device.

1. Take the TiO2 coated piece of glass and drop the fruit or plant extract with a dropper onto the TiO2. Be sure that the glass is completely covered. The white TiO2 paste should turn completely dark so there is no white left. The darker the better.
2. While you wait, take your other piece of FTO glass and find the conductive side as in step 1. Set the conductive side face up and use the pencil to coat the entire surface with graphite (pencil lead).
3. It can be hard to see but as long as you colored there should be graphite on the surface. Set it aside but keep the conductive side face up.
4. Set the dye-based glass onto a paper towel and very gently dab it with the towel to dry it off. DO NOT WIPE the glass as the TiO2 coating will come off.
5. Place a very, very tiny drop of the the iodide/triiodide (I-/I3-) electrolyte solution on the TiO2 electrode. A very small amount should be sufficient.
6. Take the two pieces of glass and assemble them into a sandwich with the two conductive and coated sides facing in. Think of a PB&J sandwich: the coated sides face in. The colored area of the glass should turn darker as it is filled with the electrolyte. Slide the graphite glass out so that its edge aligns with the beginning of the purple TiO2 coating on the other piece. Then using binder clips, clip together the two sides of the glass that are not offset. If there are any spots that don’t get coated, try removing a binder clip and then clipping it back on to move the liquid around. If that doesn’t fix it, add a little more electrolyte to the seam.

Testing the solar cells

1. The DSSC is ready to generate electricity (hopefully). All it needs is some light and a way to measure its output. You can use a multimeter to see if your solar cell is working.
2. To test your solar cell, clip one end of an alligator clip to one of the overhanging pieces of glass. Clip the other end of the alligator clip to one of the metal multimeter probes. Use the second alligator clip and clip it to the other piece of overhanging glass and the other multimeter probe.
3. Switch the multimeter setting to DCV (Direct Current Voltage) to measure the voltage of the DSSC. The 2000m setting is usually sufficient to measure the output in millivolts. An average reading in full sunlight is around 350 mV.
4. If the reading is negative, this just means the meter is measuring electricity flowing in the opposite direction. Simply switch which electrode the alligator clips are attached to and the reading will become positive.
5. Flipping the DSSC over so the dye is closer to the light can sometime increase voltage dramatically.
6. Then switch the multimeter setting to DCA (Direct Current Amperage) to measure the current. The setting of 2000u is usually sufficient to measure the current output. A typical reading in full sunlight is about 700 μA.
7. Finally, the voltage and current readings can be multiplied together to obtain the overall power of the cell. Power is defined as follows: P = current*voltage = I*V. Be sure to convert the voltage from mV to V and uA to A before multiplying.

Generating the IV curve

Terrestrial solar irradiation typically has an intensity of ~90-100 mW/cm². Indoor lighting sources can be calibrated to this intensity by changing the distance of the cell from the light source to match the current values obtained when exposed to outdoor irradiation (8). For my class, we use a halogen work lamp, but any light could be used, including sunlight.

(1) If the lamp is off, make sure you turn it on and allow it to warm up for 10 minutes.

(2) Use the light meter to determine the distance from the lamp that is consistent with 100 mW/cm². Make sure you record this distance and all measurements in this lab. This is the part that trips up people. If you use a light meter, the measurements are in lux. How do you convert from lux to mW/cm²? It isn’t straight forward. The challenge is that the conversion depends on the spectra of the light source. In our class, we estimate this using published data. It can be a bit hard to find for many lamps, so, if you’re really stuck, use the values for sunlight, but remember that this is an estimate that may be a modest approximation since the spectra for sunlight could be quite different. It at least gives one a relative starting place.

(3) Make sure you set up the circuit like figure 4.

(4) Under illumination, set the resistance at the maximum value and record the voltage. Decrease the resistance and record the voltage once again. Repeat this procedure until the voltage reads nearly zero. The current can be recorded for each corresponding point by using Ohm’s law:

where the current (i) equals the voltage (V) divided by the resistance (R).

(5) Ultimately, you’ll plot up the data.

A typical i-V curve is shown in Figure 5. From here we can use this data to determine the power conversion efficiency (h) of the cell. The power (P) can be determined by developing a spreadsheet or table using the following relationship:

P(Watt) = i(Amps) x V(volt)

Once the table is developed, P_m or maximum power is the maximum of the data set. This is the ‘knee’ of the graph in Figure 5. The corresponding current and voltage to the maximum power value are often assigned the nomenclature of i_m and V_m, respectively. Moreover, on the curve in Figure 5, the current value of zero potential is known as the short circuit current, or i_SC, while the potential of the cell at zero current is known as the open circuit potential (V_OC). These parameters are used to determine the fill factor (ff or FF) of the cell, which is a lumped parameter, and is a theoretical measure (ratio) of how much power that can be extracted from the photocell. Further, these measurements also provide insight into recombination of electron-hole pairs at the electrode-electrolyte interface as well as ohmic losses due to poor cell construction. The ff can be calculated by the expression:

The power conversion efficiency is a quantitative measurement of the effectiveness with which a solar cell converts solar energy to electrical energy. Silicon solar cells, often found atop of buildings and in solar harvesting fields, typically operate with an η value of 20-25%(10) while DSSC have an η value of ~11%.(4) So far, the best-recorded value for a solar cell in a laboratory setting is 42.3% using InGaP/GaAs/InGaAs under concentrated solar light (over 400x the terrestrial solar irradiation) (11). The efficiency value is determined by the following relationship,

where P_I is the incident light power which the cell is exposed to, often expressed in units of light intensity (mW/cm²). It should be noted that the nomenclature used in the literature may differ slightly from source to source, but still conveys the same information. If the incident light power is not readily known/determined, the ff can be used as a metric for cell performance.

These equations give rise to a method to qualitatively as well as quantitatively measure the performance of a solar cell. These metrics are often used in academic research as well in industry to determine solar cell performance.

Cleaning up

1. Juices can go down the sink. Paper towels go in the trash. You can choose to toss plastic pipettes or rinse and reuse them for future classes.
2. Save the FTO glass and simply wipe clean with water and a towel. Graphite is easily removed by a rubber eraser.
3. All other parts are reusable and should be packed away for future use.

Things to consider with these experiments

There are a host of questions that can be answered by making and testing different kinds of solar cells at home. What kinds of cells make sense under what conditions? What are the most important parameters for the major applications? What are the challenges of the DSSCs? Can you come up with a new dye formulation that is cheap but gives a reasonable efficiency and generates enough power for a particular application?

Some of my students noted that DSSCs can be used to make art. Integrating art and science is a very cool direction one could take this in. The FTO glass is one of the most expensive parts. If you are interested in moving in that direction, it is worth thinking about conductive, transparent materials. There are a number of materials out there that might be able to be used. This is just a starting point.

References

1. Anunson PN, Winkler GR, Winkler JR, Parkinson BA, & Schuttlefield Christus JD (2013) Involving students in a collaborative project to help discover inexpensive, stable materials for solar photoelectrolysis. Journal of Chemical Education 90(10):1333-1340.
2. Smith YR, Crone E, & Subramanian V (2013) A Simple Photocell To Demonstrate Solar Energy Using Benign Household Ingredients. Journal of Chemical Education 90(10):1358-1361.
3. Smalley RE (2005) Future global energy prosperity: the terawatt challenge. Mrs Bulletin 30(6):412-417.
4. O’Regan B & Grätzel M (1991) A low-cost, high-efficiency solar cell based on dye-sensitized colloidal TiO2 films. Nature 353:737.
5. Kim JB, et al. (2012) Wrinkles and deep folds as photonic structures in photovoltaics. Nature Photonics 6:327.
6. Yan SG, Lyon LA, Lemon BI, Preiskorn JS, & Hupp JT (1997) Energy Conversion Chemistry: Mechanisms of Charge Transfer at Metal-Oxide Semiconductor/Solution Interfaces. Journal of Chemical Education 74(6):657.
7. Syafinar R, Gomesh N, Irwanto M, Fareq M, & Irwan YM (2015) Potential of Purple Cabbage, Coffee, Blueberry and Turmeric as Nature Based Dyes for Dye Sensitized Solar Cell (DSSC). Energy Procedia 79:799-807.
8. Smestad GP & Gratzel M (1998) Demonstrating Electron Transfer and Nanotechnology: A Natural Dye-Sensitized Nanocrystalline Energy Converter. Journal of Chemical Education 75(6):752.
9. Michael JM, Greg J, & Ian M (1994) An experiment to measure the I – V characteristics of a silicon solar cell. Physics Education 29(4):252.
10. Harper MM, et al. (2011) Transplantation of BDNF-Secreting Mesenchymal Stem Cells Provides Neuroprotection in Chronically Hypertensive Rat Eyes. Invest Ophthalmol Vis Sci 52(7):4506-4515.
11. Green MA, Emery K, Hishikawa Y, & Warta W (2011) Solar cell efficiency tables (version 37). Progress in Photovoltaics: Research and Applications 19(1):84-92.