Battery plus supercapacitor: what each is good at
A battery and a supercapacitor are not two grades of the same thing. They store charge by different physics, and that difference decides what each is good at. A battery holds a lot of energy and dislikes delivering it quickly. A capacitor holds very little energy and does not mind how fast you take it. Neither fact is a defect; the pairing exists because the two failure modes are opposite.
The physical difference, in one paragraph each
A battery stores energy chemically. Charge is held by reduction and oxidation reactions in the bulk of the electrode material: ions move into and out of the lattice, and the electrode material itself changes state. Because the whole volume of the electrode participates, the amount of energy stored per kilogram is large. Because the process involves mass transport through a solid and a chemical change at each end, it takes time, and it wears the material out.
A supercapacitor stores energy electrostatically. Charge accumulates on either side of the interface between an electrode and the electrolyte, separated by a layer of solvated ions a few nanometres thick. Nothing reacts and nothing changes phase — ions simply gather at a surface and disperse again. Because the separation is so small and the electrode surface area so large, the capacitance is enormous by the standards of a conventional capacitor. Because only the surface participates, the energy stored per kilogram is small. The physics of that layer is the subject of a separate note.
What follows from that
Every practical difference between the two is a consequence of the paragraph above rather than a separate fact to memorise.
| Lithium-ion cell | Electric double-layer capacitor | |
|---|---|---|
| Storage mechanism | Faradaic — redox in the electrode bulk | Electrostatic — ions at a surface |
| Specific energy | High — commonly quoted in the range of roughly 150–250 Wh/kg at cell level | Low — commonly quoted at roughly 5–10 Wh/kg |
| Specific power | Modest | High — typically an order of magnitude or more above a cell |
| Cycle life | Hundreds to a few thousand full cycles, depending heavily on depth and temperature | Hundreds of thousands of cycles, often quoted at a million |
| Discharge voltage | Relatively flat plateau over most of the capacity | Falls linearly as charge leaves, because V = Q/C |
| Cell voltage | Around 3.6–3.7 V nominal for common chemistries | Around 2.7 V for common organic-electrolyte devices |
| Self-discharge | Low — months | High — days |
The numerical ranges above are the figures commonly published for commercial devices. They vary substantially by chemistry, format, temperature and how the manufacturer defines the measurement, and they are given here as orders of magnitude rather than as specifications for any particular product.
The two-dial picture: energy and power are not the same quantity
Energy is how much work a store can do in total; power is how fast it can do it. A device can be excellent at one and poor at the other, and the plot that makes this visible — specific power against specific energy, both on logarithmic axes — is called a Ragone plot. Batteries and capacitors sit in different corners of it, and the gap between those corners is not a technology gap waiting to be closed. It is a consequence of storing charge in a bulk versus storing it on a surface.
This is why the question "which is better" has no answer, and why the useful question is "which does this duty cycle need". An application that draws a steady modest current for hours needs energy. An application that draws a short violent current many times an hour needs power. Most real applications need both, at different moments, which is the whole reason for pairing them.
Why pairing them is not simply adding them
Put a capacitor bank across a battery and the current divides between them according to their impedances, not according to what the designer wanted. Getting the transient into the capacitor and the sustained load out of the battery requires that the split be arranged deliberately — which in most architectures means a converter between the two, and a control strategy deciding what goes where and when.
That converter is the cost of the pairing, in money, mass, volume and its own losses, and it is why hybrid storage is not already standard practice everywhere. The honest treatment of that trade-off is in the technical note on what makes hybrid storage hard, and the questions about it I cannot yet answer are in the open-questions list — including the one that decides adoption, which is whether the trade pays for itself in a given application.
The case for the pairing, stated plainly
The argument is narrow and it is this. Ohmic loss in a cell goes as the square of the current, so a short high-current event generates heat out of all proportion to the energy it carries, and that heat is deposited in the material that ages. If the transient is taken by a device that does not age from being used that way, the cell sees a gentler current profile. The arithmetic behind that claim is set out in the note on I²R, and it is the mechanism the Hybrid Power Pack architecture is built around.
What the argument does not establish is how much longer any particular pack lasts in any particular application. That is an empirical question that depends on the duty cycle, and a test report answers only the question its protocol asked. The Hybrid Power Pack has been cycle-tested at CIRT and environmentally tested at DRDO R&DE(E); both are test reports, not certifications, a distinction explained in how to read a test report.
When the pairing is the wrong answer
- When the load is steady. If there is no transient to buffer, the capacitor bank is mass and cost doing nothing.
- When mass or volume is the binding constraint and cell life is not. The capacitor adds the former to protect the latter.
- When the cells are cheap and easily replaced. The value of the buffer is the replacement it defers; if replacement is inexpensive, so is the value.
- When a different chemistry already tolerates the duty cycle. Some cells are built for high rate, and buying rate capability in the cell can be simpler than buying it in a second device.
Publishing that list costs nothing and is worth more than another paragraph of advocacy. An engineer evaluating an architecture needs to know where it stops working before the parts where it does.