I've heard it suggested that "solid tantalum" capacitors are dangerous and may cause fire, may fail short circuit and are fatally sensitive to even very short over voltage spikes.
Are tantalum capacitors reliable?
Are they safe for use in general circuits and new designs?
Summary:
"When used properly" tantalum capacitors are highly reliable.
They have the advantage of high capacitance per volume and good decoupling characteristics due to relatively low internal resistance and low inductance compared to traditional alternatives such as aluminum wet electrolytic capacitors.
The 'catch' is in the qualifier "when used properly".
Tantalum capacitors have a failure mode which can be triggered by voltage spikes only 'slightly more' than their rated value. When used in circuits that can provide substantial energy to the capacitor failure can lead to thermal run-away with flame and explosion of the capacitor and low resistance short-circuiting of the capacitor terminals.
To be "safe" the circuits they are used in need to be guaranteed to have been rigorously designed and the design assumptions need to be met. This 'does not always happen'.
Tantalum capacitors are 'safe enough' in the hands of genuine experts, or in undemanding circuits, and their advantages make them attractive. Alternatives such as "solid aluminum" capacitors have similar advantages and lack the catastrophic failure mode.
Many modern tantalum capacitors have built in protection mechanisms which implement fusing of various sorts, which is designed to disconnect the capacitor from its terminals when it fails and to limit PCB charring in most cases.
If 'when', 'limit' and 'most' are acceptable design criteria and/or you are a design expert and your factory always gets everything right and your application environment is always well understood, then tantalum capacitors may be a good choice for you.
Longer:
Solid Tantalum capacitors are potentially disasters waiting to happen.
Rigorous design and implementation that guarantees that their requirements are met can produce highly reliable designs. If your real world situations are always guaranteed to not have out of spec exceptions then tantalum caps may work well for you, too.
Some modern tantalum capacitors have failure mitigation (as opposed to prevention) mechanisms built in. In a comment on another stack exchange question Spehro notes:
Strangely, I can find nothing about the "ignition failure" feature in their other data sheets.
Solid Tantalum electrolytic capacitors have traditionally had a failure mode which makes their use questionable in high energy circuits that cannot be or have not been rigorously designed to eliminate any prospect of the applied voltage exceeding the rated voltage by more than a small percentage.
Tantalum caps are typically made by sintering tantalum granules together to form a continuous whole with an immense surface area per volume and then forming a thin dielectric layer over the outer surface by a chemical process. Here "thin" takes on a new meaning - the layer is thick enough to avoid breakdown at rated voltage - and thin enough that it will be punched through by voltages not vastly in excess of rated voltage. For an eg 10 V rated cap, operation with say 15V spikes applied can be right up there with playing Russian Roulette. Unlike Al wet electrolytic caps which tend to self heal when the oxide layer is punctured, tantalum tends not to heal. Small amounts of energy may lead to localised damage and removal of the conduction path. Where the circuit providing energy to the cap is able to provide substantial energy the cap is able to offer a correspondingly low resistance short and a battle begins. This can lead to smell, smoke, flame, noise and explosion. I've seen all these happen sequentially in a single failure. First there was a puzzling bad smell for perhaps 30 seconds. Then a loud shrieking noise, then a jet of flame for perhaps 5 seconds with gratifying wooshing sound and then an impressive explosion. Not all failures are so sensorily satisfying.
Where the complete absence of overvoltage high energy spikes could not be guaranteed, which would be the case in many if not most power supply circuits, use of tantalum solid electrolytic caps would be a good source of service (or fire department) calls. Based on Spehro's reference, Kemet may have removed the more exciting aspects of such failures. They still warn against minimal overvoltages.
Some real world failures:
Wikipedia - tantalum capacitors
Kemet - application notes for tantalum capacitors
AVX - voltage derating rules for solid tantalum and niobium capacitors
With the recent introduction of niobium and niobium oxide capacitor technologies, the derating discussion has been extended to these capacitor families also.
Vishay - solid tantalum capacitor FAQ
A. The 893D series was designed to operate in high-current applications (> 10 A) and employs an “electronic” fusing mechanism. ... The 893D fuse will not “open” below 2 A because the I2R is below the energy required to activate the fuse. Between 2 and 3 A, the fuse will eventually activate, but some capacitor and circuit board “charring” may occur. In summary, 893D capacitors are ideal for high-current circuits where capacitor “failure” can cause system failure.
Type 893D capacitors will prevent capacitor or circuit board “charring” and usually prevent any circuit interruption that can be associated with capacitor failure. A “shorted” capacitor across the power source can cause current and/or voltage transients that can trigger system shutdown. The 893D fuse activation time is sufficiently fast in most instances to eliminate excessive current drain or voltage swings.
Capacitor guide - tantalum capacitors
I did some Internet research on tantalum-capacitor failures and found that the tantalum capacitors' pellets contain minor defects that must be cleared during manufacturing. In this process, the voltage is increased gradually through a resistor to the rated voltage plus a guard-band. The series resistor prevents uncontrolled thermal runaway from destroying the pellet. I also learned that soldering PCBs at high temperatures during manufacturing causes stresses that may cause microfractures inside the pellet. These microfractures may in turn lead to failure in low-impedance applications. The microfractures also reduce the device's voltage rating so that failure analysis will indicate classic overvoltage failure. ...
Related:
AVX - surge in solid tantalum capacitors
Failure modes and mechanisms in solid tantalum capacitors - Sprague / IEEE abstract only. - OLD 1963.
AVX - FAILURE MODES OF TANTALUM CAPACITORS MADE BY DIFFERENT TECHNOLOGIES - Age ? - about 2001?
Effect of Moisture on Characteristics of Surface Mount Solid Tantalum Capacitors - NASA with AVX assistance - about 2002?
Hearst - How to spot counterfeit components
Sometimes it's easy :-) :
Added 1/2016:
Related:
Test for reverse polarity for standard wet-aluminium metal can capacitors.
Brief:
For correct polarity can potential is ~= ground.
For reverse polarity can potential is a significant percentage of applied voltage.
A very reliable test in my experience.
Longer:
For standard wet Al caps I long ago discovered a test for reverse insertion which I've not ever seen mentioned elsewhere but is probably well enough known. This works for caps which have the metal can accessible for testing - most have a convenient clear spot at top center due to the way the sleeve is added.
Power up circuit and measure voltages from ground to can of each cap. This is a very quick test with a volt-meter - -ve lead grounded and zip around cans.
Caps of correct polarity have can almost at ground.
Caps of reverse polarity have cans at some fraction of supply - maybe ~~~= 50%.
Works reliably in my experience.
You can usually check using can markings but this depends on intended orientation being known and clear. While that is usually consistent in a good design this is never certain.
With the advent of compact inexpensive high-value (10uF and beyond, rated at 6.3, 10, 16V and so on) X5R and X7R (reasonable dielectrics) ceramic capacitors there seems to be much less reason to consider tantalum capacitors.
One of the differences is that tantalum caps have an ESR that is of the order of ohms. On some LDO regulators, that's an advantage, in that the LDO won't oscillate like a banshee. In such cases, I'd prefer to use a ceramic capacitor and a series resistor.
On some sensitive analog circuits, I think there may be an advantage to tantalums over ceramic caps in reduced microphonics (in ceramic caps, due to piezo-electric activity).
One guideline in using them : if the current through the cap is strictly limited in the event of a failure, go ahead.
Limited to what? I would suggest 0.1A. I would feel wary of using them to decouple a 1A or higher supply rail, and would not personally use them across a 10A supply. (Been there, seen the fireworks; Russell's pictures do not exaggerate.) I have to say I have no hard evidence of a truly "safe" current and comment on these figures would be welcome.
But many supplies or bias voltages in analog circuitry have relatively high source impedances or strictly limited currents, and I would use them there.
EDIT based on new (to me!) information ...
At least one manufacturer is offering Niobium Oxide capacitors in very similar packaging and range of values and voltages. In what might be read as a tacit admission of tantalum's problems described here, the datasheet contains the statement "Failed OxiCap® will not burn up to category voltage" and a cute little logo...
[Disclaimer : I have neither used these capacitors nor attempted to verify the claim!]
A short note on "why Tantalum instead of large MLCCs":
MLCCs with X5R and similar dielectrics are characterized at 0V bias. However, when running at e.g. 100% of rated voltage the effective differential capacity may only be 10% of the rated one (!). Especially very small caps with high voltage rating show a dramatic decrease in capacity when biased.
Example 1: 0402 MLCC, X5R, 10µF, 6.3V: 3.5µF left at about 3V.
Example 2: 0402 MLCC, X5R, 2.2µF, 25V: 1.0µF (!) left at about 3V.
That data is well shown in the online-datasheets from TDK.
why Tantalum instead of large MLCCs. It should be posted there, not in an unrelated question.
\$\endgroup\$
Some additional stuff from my side:
Yes, it can be stated that Tantalum Caps are safe.
They are not only used in the "rough" environment of consumer portable devices (notebook, smartphone - I never heard of a fire in a smartphone due to the caps) but are also utilized in medical implants such as heart pacemakers, cochlear implants or spinal cord stimulators.
With respect to reliability, the operating voltage has the strongest impact (much more than the temperature). The acceleration factor is
AF=exp{(V/VR-1)*18.772} according to the following NASA document: https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/20110015254.pdf
For medical implants, the derating proposed by e.g. Vishay is 40% (so you would use a 16V cap for 10V or 10V rating for 6V applications). According to the upper formula the increase in lifetime is a factor of 1140.
Pls. do always keep in mind that there is no system which will not fail: The only question is the time for the cumulative error. I made my Master Thesis at Infineon. I think I can remember that the MOSFETs within safety-critical automotive systems had an allowed failure rate of 10ppm within 10.000hrs when operated on max. conditions (Temp & voltage)
There may be space-limited applications where tans are better, but that's about all. I avoid tans if I can. Common parts fail by letting the smoke out. They don't like high turn-on currents surges making them a poor choice for most power supply filtering. At least use the highest voltage part you can. They don't like high humidity which can hurt self-healing. Ceramics have gotten better and can replace them in many applications, as can aluminums some times.
