For a high differential gain, compressively strained InGaAs quantum wells (QWs) are commonly employed in the wavelength range from 850 to 1200 nm [
11–
18,
20–
24]. Shallow surface etching has been adopted to tune the photon lifetime to enhance the modulation bandwidth but introduces mirror losses as shown in Fig. 3 [
36]. However, Bimberg’s group at TU Berlin experimentally showed that a short laser cavity, for example,
l/2-cavity, can greatly enhance the confinement factor to 5.1% from 3% in a 3
l/2-cavity without degrading other parameters [
37,
38]. Figure 4 shows that there is a maximum peak of the field intensity distribution at the QWs region for the
l/2-cavity compared with the 3
l/2-cavity. At the same time, a short cavity means a low photon lifetime resulting in a low
K-factor [
37]. The reported highest speed of 71 Gbps for VCSELs was achieved with a
l/2-cavity [
17]. For the energy efficiency, Bimberg’s group reported the record lowest energy efficiency of 56 fJ/bit dissipated energy with a 3.5-µm VCSEL, as shown in Fig. 5 [
28]. It has been shown that a small oxide aperture help to achieve a high energy efficiency because of the single mode operation and large mode spacing [
28,
33,
34]. To improve the temperature stability, the wavelength of the laser cavity resonance is detuned from the gain peak wavelength in the laser design [
39]. The gain peak has a shift rate of 0.396 nm/K for In
0.21Ga
0.79As/GaAs
0.88P
0.12 QW system, whereas the cavity resonance wavelength shifts to longer wavelengths with a rate of 0.061 nm/K. Thus the cavity resonance wavelength is positioned at the longer wavelength side of the gain curve at room temperature in the design, as shown in Fig. 6(a). Figure 6(b) shows that at 85°C error-free data transmission at 46 Gbit/s has been achieved for 980-nm VCSELs [
31], and 50 Gb/s error-free data transmission at 90°C for 850-nm VCSELs [
32].