There are two regulatory channel plans that come into view when discussing 802.11 channels: The ISM Bands and the U-NII bands. In 5 GHz (802.11n, 802.11ac, 802.11ax) the U-NII band definitions overlap with those defined in the ISM definitions. ISM definitions were created by the FCC in the United States. U-NII definitions are International in nature. The ISM frequencies and most of the U-NII frequency allocations are unlicensed. Devices communicating in unlicensed frequency bands have no regulatory protection from interference from other devices transmitting in the band. This is the nature of all unlicensed operations. The breakdown is as follows:

DFS (Dynamic Frequency Selection) is an FCC-mandated requirement for equipment operating in the 5.250-5.350 GHz and 5.470-5.725 GHz U-NII bands (as might be used by 802.11n, 802.11ac, and 802.11ax). Weather radar, military radar, and some front-facing aircraft weather radar may operate using frequencies in these U-NII bands. Specifically, the U-NII Middle Bands (U-NII-2A and U-NII-2C, Channels 120-128). In 2007 the FCC introduced regulations requiring that devices operating in these frequency bands automatically detect the presence of radar signals and switch to alternate channels if radar is present. Certified 802.11 devices are able to take advantage of the additional channels in the U-NII Middle range with the caveat that if radar is detected then the access point is going to immediately update their radio settings and move to another channel. Coupled with the, 802.11 5 GHz access points implement a feature called Transmit Power Control (TPC) on every DFS channel so as to minimize the output power in deference to potential radar users of the band. The IEEE 802.11h standard specifies the DFS and TPC requirements for 5 GHz 802.11 operation.

It may be counter-intuitive that channel numbers aren't always numerically consecutive. This is because the underlying channel numbering scheme increments for every 5 MHz in frequency change. 

802.11 WiFi access points and WiFi client devices that are operating on the same channel and which are within range of each other (that is, the can receive each other's transmissions) operate using a "listen before you speak" rule for taking turns transmitting. This rule is based on performing a Clear Channel Assessment (CCA). When a WiFi device (access point or client device) detects that something is currently being transmitted it defers transmission, waits a random period of time (albeit a very short, random period of time) and returns to the "listen" phase of the CCA process. One important underlying reality regarding CCA is that all of the devices that can detect each other's transmissions on a particular channel take turns and, therefore, they effectively share the available channel capacity. For example, consider a 150 Mbps connection rate (perhaps 2-stream MIMO at 72.5 Mbps modulation rate). With a 150 Mbps connection rate it is reasonable to assume that something close to 130 Mbps of actual data throughput can be obtained. In this case, all the clients on a channel (that can "hear" each other on the channel) share the 130 Mbps of actual throughput. If there are 10 client devices simultaneously attempting to transmit data they will each perceive that the channel gives them a capacity of 13 Mbps throughput.

When a single frequency sine wave is modulated (altered with a pattern to represent bits) the resulting waveform requires additional frequencies, close to the carrier frequency, to accommodate the modulation process. As modulation becomes more complex and changes to the sine wave carrier occur more quickly the required additional frequencies increase. This is the reason that a WiFi channel occupies more than just a single frequency. Each WiFi channel is 20 MHz wide to accommodate the "upper side band" and "lower side band" portions of the modulated carrier frequency and there is an additional 1 MHz allocation on either side of the 20 MHz channel making each channel's total width 22 MHz.

The 802.11 standard defines fourteen channels for use by 2.4 GHz 802.11 WiFi. Only channels 1 through 11 are available for use in the United States. Each channel is 20 MHz wide with an additional 1 Mhz on the low and high end of the channel for inter-channel spacing. This results in each channel requiring 22 MHz of bandwidth. The channels are spaced 5 MHz apart. Because of this spacing it is not possible to actually use 11 channels in the same location since the 22 MHz width would overlap and cause communication failure. The consequence is that the 2.4 GHz 802.11 band is considered to have 3 usable, non-overlapping channels: 1, 6, and 11. Some implementations may try to take advantage of a 4 channel approach (1, 4, 7, and 11) with the hope that the overlap will not cause severe problems. The more effective channel overlap solution is to simply move to a 5 GHz 802.11n, 802.11ac, or 802.11ax technology. 

Best-in-class commercial-grade WiFi equipment operates both in the backward-compatible, legacy 2.4 GHz spectrum (using dual-radios) but optimal capacity and throughput, providing the best end-user experience, occurs in the 802.11n, 802.11ac, and 802.11ax 5 GHz frequency bands. Like the 2.4 GHz WiFi channels, each 5 GHz channel is 5 MHz wide. There are 23 channels specified for 802.11 WiFi, however, unlike the 2.4 GHz channel definitions, only the non-overlapping channel numbers are used for device configuration. Consequently, a list of 5 GHz 802.11n, 802.11ac, or 802.11ax channels is non-contiguous (with the otherwise overlapping 5 MHz-wide channel numbers ignored). The channels are:

Adjacent 802.11n, 802.11ac, and 802.11ax 20 MHz channels can be "bonded" together to create a wider transmission channel. While this is true for 2.4 GHz 802.11n it is impractical for any commercial WiFi network since it would allow for only 1 40 MHz channel in the 2.4 GHz Channel 1 through Channel 11 frequency space. It should be noted that the standards allow for "non-contiguous channel bonding" however it's unclear if any commercial equipment actually supports this arrangement.

When a 40 MHz channel is configured it actually has slightly more than twice the capacity of the original two adjacent 20 MHz channels. This is because the slight gap in frequency that's allocated for adjacent channels now becomes part of the single, double-wide channel. If an 802.11n transmitter is operating in a 20 MHz channel and can establish a 72.2 Mbps connection then a 40 MHz channel would provide for a 150 Mbps connection. When you double the channel width you double (plus about 4%) the capacity of the resultant double-wide 802.11n, 802.11ac, or 802.11ax channel.

When a wireless LAN uses 802.11n, 802.11ac, or 802.11ax exclusively it's referred to as an "802.11 Greenfield Mode" implementation. Most wireless LANs will be required to provide support for 802.11b/g clients in the 2.4 GHz ISM band using conventional 20 MHz channels. Moreover, not all 802.11n, 802.11ac, or 802.11ax devices will necessarialy support wider than 20 MHz channels. If a device were to transmit in a 20 MHz channel which was half of a double-wide, bonded 40 Mhz channel there would be a problem. In mixed-mode networks (20 MHz channel width devices in the presence of 40 MHz devices, etc.) the only 40 MHz channels that can be successfully used are those where both of the 20 MHz channels where a 20 MHz transmitter is not currently active. To meet this requirement, a transmitter using a 40 Mhz (or 80 MHz, or 160 MHz) wide channel will first broadcast, in each of the bonded 20 MHz channels, it's intention to use the channel (a RTS/CTS mechanism). In a non-greenfield mode deployment this does introduce some degradation and prevents attaining maximum throughput from the bonded channels.