2022-03-03 - By Robert Elder
If you ask a software developer the question 'What are some example uses of semiconductors?', they'll probably mention 'silicon chips', like the ones found in processors or graphics cards. Indeed, this is probably the answer that I would have given just a couple years ago, but I have since become enlightened as to the true power of semiconductors. In this article, I will provide some familiar (and some surprising) photographic examples of semiconductors that we use in our everyday lives.
Everybody knows that CPUs have 'transistors' in them, but have you ever actually seen a transistor? Here's a single transistor in what's called a 'TO-92 package':
A transistor is a general-purpose component that can be used to switch or amplify electrical signals:
Here's what it looks like if you smash it open:
And here's a closeup of what appears to be a small piece of silicon inside the transistor package:
You've probably used a 'USB flash drive', a 'microSD card', or an 'SD card' like this one before:
Here's what it looks like when you smash one open:
And here's what the packaged 'chip' look like if you smash it open even more. Take note of the shiny silvery-coloured parts:
Here's an image of an entire intact silicon die of a flash memory chip that was never covered inside an epoxy package like the one above. This particular flash memory chip is not the same type of chip used in the blue SD card above. This particular flash memory chip is for 2GiB of flash storage:
The bright colours are from thin-film interference within the thin oxide layers on the surface of the chip. The edges of the chip contains lots of details that encode the logic for reading and writing data to the chip:
The highly uniform area in the center of the chip contains features that actually store the data:
Here is a closeup of the memory cells (along with some dust particles):
LED Lights (Light Emitting Diodes)
LEDs are also made of semiconductor materials, such as gallium arsenide, indium gallium nitride, gallium phosphide and many others. The exact choice of material is an important factor in determining which wavelengths of light will be emitted by the LED. In particular, different materials have different 'band-gap' values, which is an important factor in determining the energies (and therefore wavelength) of emitted photos.
When people talk about harnessing the power of 'solar energy', they're talking about using 'solar cells' to convert light into electricity. These solar cells also happen to be made of semiconductors. The most common semiconductor material used to make solar cells is silicon:
Some high-efficiency solar cells use cadmium telluride instead. Silicon solar cells can be purchased in a number of different forms according to the various crystal structures of silicon, such as monosilicon, polysilicon, and amorphous. Each form comes with its own pros and cons in terms efficiency and production cost. Other factors like doping concentration, impurities, crystal grain size and geometry can affect the efficiency and cost. This leads to a surprisingly large amount of research that can be done on this topic.
Here is a demonstration of measuring the open-circuit voltage of an individual solar cell:
Most people don't think of power supplies as a 'computing device', but that doesn't mean that they don't contain any semiconductors!
Here's an old power supply from one of my servers:
And if you open it up, you'll see all kinds of non-semiconductor materials, like heatsinks, inductors, capacitors etc:
As technology improves, the trend has been to increase the efficiency of power supplies like these. This requires an increase in the complexity of the semiconductor devices that they use. You can find many examples of these by reviewing the topic of Switched-mode Power Supply design.
Here is a DS18B20 thermal probe. The metal tip is the part that actually senses the temperature:
Here's what the metal tip looks like if you cut it open:
Take note of the silvery slab of material that rests on top of the copper in the center of the image above. That looks like a semiconductor to me!
Here is my video camera with the lens removed:
It uses a CMOS (Complementary Metal Oxide Semiconductor) image sensor to detect light using an array of tiny photodetectors:
The photodetectors are made to be sensitive to different colours and arranged in a Bayer filter pattern so that full colour information can be reconstructed to produce an image.
Photodiodes are useful devices for detecting light levels. As their name implies they are also diodes:
They work a lot like mini solar cells, although their exact chemistry makes them better suited for measuring light levels instead of achieving maximum energy collection:
However, you can still put a few of them in series and collect meaningful amounts of energy with them:
For example, I was able to use the above photodiodes to charge up a 2200uF capacitor to just over 3 volts after about 45 minutes in a fairly dim room in the middle of winter. This represents a total charge stored in the capacitor of about 0.01 Joules, which is enough to briefly flash this 0.5 Watt LED light:
'Chips' Doesn't Just Mean 'CPUs'!
People are more aware of things that they can see, so it's no surprise that people are more aware of the 'chips' that run graphical operating systems in our laptops, phones and desktop computers. Chips that are used in non-graphical operating systems (like servers, cloud systems, industrial automation, etc.) don't get as much attention, but they are actually more numerous than the 'chips' that run our personal computers.
Therefore, it should come as no surprise that small special-purpose chips that don't even run operating systems at all are orders of magnitude more numerous than any kind of 'processor'. However, these kinds of chips are effectively invisible to anyone who doesn't regularly work with embedded systems.
This is where the term Microcontroller unit, or 'MCU' comes in. There are many qualitative differences between a 'processor/microprocessor' and a 'microcontroller', but I'll try to provide my own brief definition of my own: A 'microcontroller' differs from a 'processor' by providing a more bare bones and raw level of access to the hardware being controlled than a processor would. Typically, software written for a 'processor' runs within a software process in the context of an operating system and will most likely be subject to process-isolation features like CPU usage limits, limits on privileged instruction execution, and access controls on memory. In contrast, most software written to run on a microcontroller IS the operating system, and will have the ability to use 100% of the CPU, execute any instruction it wants, and read/write to any part of memory. Furthermore, a general-purpose 'processor' might prefer to offload I/O operations, interrupts, or interactions with peripherals to a separate chip, but a 'microcontroller' is likely to have all these functions integrated onto the same physical chip. The overall benefits of a microcontroller are lower cost and greater simplicity, but the drawback is that most of them are only designed to work in one very specific application to control very specific hardware.
For example, here is a chip that serves as an 8-bit multichannel sound controller, with a clock speed of 8 Mhz. It offers a serial peripheral master interface, 1 megabyte of ROM, and a real-time clock. It's listed purposes include audio processing, sound mixing, PCM wave processing, speech and melody processing functions:
And here is a more 'general-purpose' 8-bit microcontroller with 17 interrupt sources, 22 I/O pins, an 8-bit UART, and a programmable watchdog timer. One of its product specification sheets lists the following potential applications: "Portable devices, Digital information appliances (broadband/OA equipment), Digital consumer (DSC/DVC/DVD/DTV/STB), Automotive equipment, AudioTV/communications (cordless phones), electric appliances, inverter appliances".
This chip was made specifically for a 900Mhz cordless phone, and features a phase locked loop, an infrared detector, and a compander:
The above chip may do a great job inside a cordless phone, but it probably won't have much use in your personal laptop.
Chip Interconnections Vs. PCB Traces
The more you read about how chips are made, the more you realize that they're just miniaturized versions of other macroscopic circuits. That's probably why chips are often called 'integrated circuits'. If you've ever taken apart an electronic device and seen the traces on a green 'printed circuit board', you can make a direct comparison to the copper interconnects inside of a 'chip':
'Wires' Inside A 'Chip' (Integrated Circuit)
'Wires' Inside A Printed Circuit Board'
Pretty much any circuit that you can imagine wiring up on a breadboard, or soldering onto a PCB, you can probably also turn into an integrated circuit (with a few practical limitations for things like huge capacitors/inductors, moving parts etc.). That's why 'chips' have become so popular!
One of the common themes with most of the semiconductors we've discussed so far is that they use silicon. It's worth mentioning that silicon comes in a number of different crystal forms. For example, here is a piece of monocrystalline silicon that has a completely uniform crystal lattice, completely free of any grain boundaries:
And here is a piece of polycrystalline silicon with many disorganized grain boundaries:
Here is a piece of dendritic silicon that has a branching crystal structure:
Silicon isn't the only semiconductor! Germanium is a also commonly used too. Here is a crystal of pure germanium:
For certain applications, germanium is preferable to silicon due to differences in its material properties. For example, silicon diodes typically have a forward voltage of approximately 0.7 volts, whereas germanium diodes have a forward of approximately 0.3 volts.
Germanium is also commonly used to make infrared optics. Here is an example of an infrared lens:
Fundamentals Of Semiconductors & Further Reading
So far, we've only scratched the surface of what semiconductors can do. In order to imagine what more they can do for us, it's useful to identify what properties of semiconductors make them so varied and useful:
1) Band Gaps: The concept of a band gap is the most important concept to understand about semiconductor materials. The 'band gap' is a number that quantifies the 'energy gap' required to move an electron up from the valence band to the conduction band. In fact, it's helpful the think of all materials (including non-semiconductors) as having a 'band gap', but for metals the electrons are already in the conduction band, and for insulators the electrons require a HUGE amount of energy to get into the conduction band. In reality, the concept of 'conductors' and 'insulators' that we learn about in grade school is really just a 'close enough' bit of pseudoscience that is used to avoid discussing quantum mechanical properties of electrons in solid state physics.
For things like solar cells, photodiodes, or LEDs, the numerical value of the band-gap is important because it determines what energy levels (wavelengths) a photon will have when it is emitted or absorbed by an atom of that semiconductor material. Note that the 'energy' of a photon determines its wavelength (colour).
For things like diodes or transistors, the band-gap is one of many factors that will influence the forward voltage of the device. With all other factors being equal, a higher band-gap material will require a higher forward voltage, but have less leakage current. A lower band-gap material will have a lower forward voltage, but high leakage current.
The function of a semiconductor device also depends a lot on whether the material has a direct or indirect band gaps, but I won't discuss that at all in this article.
2) Doping Dependent Resistance: When discussing the properties of semiconductors, the simplest scenario to consider is a pure sample of that semiconductor substance. However, the actual resistance of a semiconductor also depends on the concentrations of any impurities that may be mixed in with it. Even small concentrations like 1 part per million can make a significant difference. The impurities are called dopants. Two of the most common dopants found in silicon semiconductor devices are boron and phosphorus. As mentioned in the previous section, the band-gap properties of a semiconductor are extremely important and the process of doping the semiconductor will also have an effect on the band-gap of the material.
3) Temperature Dependent Resistance: Temperature also has a significant effect on the electrical properties of a semi-conductor. In particular, the electrical resistance of a semiconductor decreases when the temperature increases. For typical 'conductors' their electrical resistance increases when the temperature increases. Here are some real-world measurements that I made to show the changes in electrical resistance of copper wire versus a chunk silicon as the temperature increases:
This sensitivity to temperature can be a desirable property if you want to build a semiconductor device that measures temperature: You would start by characterizing the temperature/resistance graph of a chunk of some semiconductor material. Then, to take a temperature measurement, infer the resistance/actual temperature by measuring the voltage drop across your chunk of semiconductor material.
The effect of decreased electrical resistance with increasing temperature can also be an undesirable property of a semiconductor: A common problem that is experienced with semiconductor devices (in particular, with LEDs lights), is thermal runaway. As the semiconductor device operates, it draws current, some of which is dissipated as waste heat in the device itself. As mentioned previously, increasing the temperature of a semiconductor causes a decreases in its electrical resistance, but a lower resistance means a higher current draw. Since the heat dissipated in a resistor is based on the square of the current, the device can experience thermal runaway if the waste heat from the device does not dissipate to the surroundings fast enough. As more and more heat builds up in the device, it continues to draw more and more current heating up even faster until either the device melts, or it reaches an equilibrium heat dissipation rate with the surroudings.
There are also two extremely important temperature-related semiconductor devices that I did not have time to include elsewhere above: Peltier devices, which turn electricity into a temperature gradient (for heating or cooling) and Seebeck generators that turn a temperature gradient into electricity (for energy harvesting).
4) Crystal Grain Structure:
As we've seen above, semiconductors can be organized into different crystal structures. Individual crystal grains can be large or small, and arranged in any manner. The concept of a band-gap is easy to understand when you only have one atom, or several atoms arranged in a homogeneous structure, but reality is more complicated than this. In a crystal, the electrons are able to move throughout the structure, so it is naive to just consider how electrons (and band-gaps) behave in relation to one atom in the crystal. Instead, it is necessary to consider how electrons move throughout the entire crystal, taking into consideration the crystal grain boundaries and any dopant(s) that may be present. If you want to read more about this topic, see solid-state physics.
5) Crystal Anisotropy:
If things weren't complicated enough already, it's now time to introduce the word 'anisotropic' which the dictionary defines as follows: 'in physics, the quality of exhibiting properties with different values when measured along axes in different directions'. If we consider a pure single-grain silicon monocrystal with no doping at all, the arrangement of the atoms on the surface of the crystal will look different depending upon which direction you approach the crystal from:
Image used with permission from: https://www.microchemicals.com/technical_information/silicon_crystallography.pdf. Accessed January 10, 2022: md5:6338a2b3d866a49474b5ebb3686cf0ec silicon_crystallography.pdf
These crystal faces are usually identified according to their miller index.
Silicon has a diamond cubic crystal structure which gives it this anisotropic property. This is important because it turns out that important material properties like electrical resistance can actually depend on the direction that you travel through the crystal lattice! This is the part where my ability to even understand this topic gets very fuzz. To the best of my knowledge, most literature on this topic seems to stop discussing the concept of a 'band-gap' of the semiconductor material, and instead uses the term 'band diagram' which seems to be a 3D generalization of the concept of a 'band-gap'. This does seem reasonable after all because the different crystal faces of silicon (shown above) are not symmetric and one would expect a free electron to interact differently with each crystal face.
As you can see, there are a lot of variables to play with when developing a semiconductor device: Intrinsic material band-gap, doping, temperature, grain structure, and crystal orientation. When looking at the humble solar cell or diode, you might wonder why we're still doing so much research on semiconductors. However, considering how many potential permutations there are of these variables to research, it is not surprising why we're still discovering new things.
How To Make A CPU - A Simple Picture Based Explanation
Jeri Ellsworth & the Robot Uprising of 2038
CPUs Are Smaller Than You Think
Can You Create A Wireless Bridge With ESP8266 Modules?
Routing Packets Directly Between ESP8266 Modules
Using The GY-511 LSM303DLHC Accelerometer To Detect Waving Motion
All About SD Card Flash Memory Corruption in Cameras & Raspberry Pis
Why Bother Subscribing?