The transistor put us on the path to semiconductor electronics research. It led to the integrated circuit, optoelectronics, light-detecting devices, diode lasers, LEDs, and now further to the transistor laser—a true laser and three-terminal photonic transistor active element.
(Counter clockwise from bottom left) Nick Holonyak Jr., Milton Feng, Gabriel Walter and Richard Chan. A light-emitting transistor is shown in the background. The bullseye white ring is the quantum-well base recombination light visible from the top with a small Bragg reflector buried in the transistor collector.
Arguably the most important transistor ever made was the Bardeen and Brattain point-contact transistor, which came in 1947. It was the original bipolar semiconductor triode (i.e., +/– charge conduction, the positive hole as important as the negative electron). It was the device that started it all, the prototype that ultimately doomed the vacuum tube. It led to the integrated circuit and a vast new branch of electronics that made large-scale computers possible.
Eventually, because of +/– bipolarity and electron-hole recombination, it also led to the diode laser and the LED. In short, the first transistor ultimately changed the world's economies and history. The point-contact transistor was primitive but ingenious—it included just the required bare minimum, a small crude "slab" of germanium (Ge). It was simply an n-type "base" crystal with an ordinary Ohmic current contact attached to the bottom (terminal #1, B), with an input metal point-contact minority carrier (hole) emitter (terminal #2, E) on the top surface, and nearby an output metal point-contact "collector" (terminal #3, C).
Indeed it was just enough—a "base" crystal and three contacts (EBC)—to reveal the three-terminal low-impedance input/high-impedance output "transfer-resistor" idea. The transistor was a radically new bipolar active amplifying and switching element. When Bardeen and Brattain demonstrated their device, which in the beginning was not known and not named, it was totally new, not previously anticipated by anyone. It emerged in physics like a strange mushroom popping out of the ground in a forest.
For what it was and what it revealed as a progenitor, the point-contact transistor of Bardeen and Brattain outranks all other transistors. Because of it, the hole in a real working sense was elevated to the importance of the electron. It ceased to be, as many had believed, just a mysterious construct of quantum physics. It was real, and it was put to work. Of further consequence, the electron and hole were connected across the gap by the photon (ph, hv ~ Eg), pointing clearly at the possibility of sooner or later a new electronics based on the electron, hole and photon—all three. After the transistor, it was bound to happen.
What many people may not know or appreciate, however, is that the very existence of the electron-hole bipolarity (–/+) set the path to today's diode laser, LED, optoelectronics â€¦ and now the transistor laser.
The point-contact transistor is not obsolete. It still teaches, still informs. Although abandoned in use for technical reasons, it remains relevant. It was quickly superseded technologically but not conceptually, with the p–n junction ("p" positive, "n" negative, hole and electron, acceptor and donor doping) replacing the point-contact technology out of World War II. Even now, the Bardeen and Brattain experiment and device also focuses attention on a largely overlooked matter—the singular role of the base in transistor operation.
Why the base? A simple, revealing answer: In the case of the point-contact transistor, only the base (n-type) was semiconductor crystal. Nevertheless, in operation, it was bipolar (h and e, +/–), even though structurally reduced to a minimum, or maybe better, it arose from a minimum, simply a crude piece of n-type crystal. No other transistor was ever so simple or more enlightening, or pointed so sharply at the essence of the transistor, to the semiconductor base, the n-type Ge "slab" itself, and how it connected the hole input and output. However, with the electron current (IB > 0) of the n-type base, it separated the forward-biased hole input (low impedance) from the reverse-biased hole output (high impedance), thus affording gain. The base and base current were essential.
Simply from Kirchhoff's law, we see that IE + IB + IC = 0; "alpha" gain by definition is IC ≡ –α IE, 0 ≤ α ≤ 1; "beta" gain is thus β ≡ ΔIC/ΔIB, with β = α/[1–α].
Electron-hole bipolarity and recombination
It is the base and what it does that makes the transistor what it is. In some respects, the base is all that matters. Base current, which is vital, is carried by electron-hole (–/+) recombination, a gift of the semiconductor and its energy gap. Note that recombination is a source of light, or heat and loss. The energy gap (Eg) of the semiconductor is everything: It separates and "connects" across the gap (ΔE = Eg) by recombination, or by photon (light) absorption (e-h generation), the conduction band electron (–) and at lower (electron) energy the valence band hole (+). The energy gap matters; the current input-output separation matters.
The transistor base, as such, connects input and output, and the base current separates them. Mostly ignored, or misunderstood, or forgotten, it is the base and base current (IB) that offers still something more from the transistor, but what? Isn't the base current just carrier recombination and loss (heat), or is it light? If light, where is it?
In the germanium energy-band structure (the quantum mechanical E–k dispersion diagram), the electron is offset in momentum from the hole (wavevector, ke ≠ kh), and in comparison a visible or near-infrared (IR) photon has little momentum (kphoton ≈ 0). For this reason, it did not make much sense for Bardeen and Brattain, and their successors working with Ge or Si (ke ≠ kh), to look for base-current electron-hole recombination to supply much light (visible or infrared). This is a consequence of the requirement for the conservation of energy and momentum (k) in e–h recombination, and, for Ge or Si, ke – kh = Δk > 0, with Δk >kphoton.
However, it is a different matter to look 50 years or more later at the recombination product, the output of the base recombination current—particularly, let us say, in the case of the world's "fastest" (highest frequency) direct-gap (ke = kh, Δk ~ 0 ~ kphoton) heterojunction bipolar transistor (HBT), which operates at extremely high current density (~ 106A/cm2) and a beta-gain β ~ 102. In other words, the base recombination-radiation current is on the order of IB ~ 104A/cm2, which, in a small confined base region, a cavity region because of reflective boundaries and thus some Q, is sufficient to support stimulated recombination and laser operation.
Is it indeed possible? The transistor base current (IB), as was the case for many years, could not be ignored or regarded as simply loss (heat). Could it, in fact, yield a laser? The HBT was the right place to look for the answer.
(Top) Energy band diagram and charge population distribution of a quantum-well (QW) base n–p–n heterojunction bipolar transistor (HBT) laser (shown with generic cavity mirrors). Note the tilted-charge population due to the zero-charge-density boundary condition imposed by the reverse-biased collector. (Bottom) Transistor laser I-V characteristics showing in the upper part the phenomenon of I-V curve-spacing compression (gain β = ΔIC/ΔIB, βstim < βspon) above lasing threshold, IB,th ~ 22 mA.
Quantum-well base three-terminal laser
It is the point-contact transistor, and how closely (focusing on the base) the high-speed HBT emulates it, that led us to consider, over numerous coffee discussions and questions about high-level device operation, the prospect of converting an HBT and its base current (IB) from a source of spontaneous recombination to a source of stimulated recombination. We pondered the possibility of whether we could, with improvements of the base-region cavity and the recombination radiation, operate a three-terminal transistor laser (TL).
From the use in Urbana of a quantum well "collector" in a diode laser in 1977, we know that an HBT base has room to include a QW "optical collector," a second "collector" put in the base region to tailor-make carrier recombination, the purpose being to reduce HBT gain (β ≡ ΔIC/ΔIB) from β ~ 100 to 0.1 < β < 10. We thus adjust and increase, via the QW, the base current IB and the recombination radiation intensity.
As a consequence, we also decrease the gain and increase the transistor bandwidth—still operating, with fixed gain-bandwidth, the transistor as a transistor. The super-high current-density high-speed operation of the HBT, approaching roughly 1012 Hz, set a path directly toward the conversion of the previously HBT base-current loss into stimulated recombination, into a microwave-speed modulatable QW transistor laser, simultaneously an electrical transistor and a unique three-terminal laser.
The transistor changed, and the laser changed. The top portion of the figure above shows an energy-band diagram of the charge population distribution of a quantum-well (QW) base n–p–n heterojunction bipolar transistor (HBT) laser. Generic cavity mirrors are included to identify the device as a transistor laser (TL). Note the tilted-charge population, owing to the zero-charge-density boundary condition imposed by the reverse-biased collector.
Signature of the transistor laser. (Top) SEM picture showing in cross-section the edge emission output (the bullseye) of a transistor laser operating at room temperature and (inset) the TL of cavity length 800 µm. (Center) The transistor laser current gain (β ≡ ΔIC/ΔIB, βstim < βspon) revealing the shift (the "step") in base operation from spontaneous recombination to stimulated at the threshold of laser operation (IB = 45 mA). The spectrum (inset) shows laser operation at λ = 1,010 nm (IB = 60 mA). (Bottom) Optical response of the 19-GHz bandwidth transistor laser (TL) with no carrier-photon resonance owing to the fast TL base lifetime τB, spon ~ 29 ps.
Electrical signature of the transistor laser
For it to be a transistor (IB > 0), the base of the laser uses recombination. Thus for a direct-gap (ke = kh), the HBT produces light ke– kh = Δk ~ 0 ~kphoton), which can be arranged to be stimulated and yield a laser. The effect of stimulated recombination in turn changes the HBT electrical properties because of the reduced base carrier lifetime (τstim < τspon). In other words, the HBT β-gain becomes βstim < βspon, which means the collector current-voltage characteristics of the TL-HBT (ordinarily quite uniform) compress at laser threshold into closer spacing—in other words, into lower β-gain, as is clear by the closer spaced (upper) curves in the lower portion of the figure above. This is the unmistakable signature (confirmed by spectral narrowing) of the TL and, immediately from the electrical characteristics alone, it signifies laser operation—something not previously known or observed for more than 50 years and not part of thousands of textbooks and references.
The point is: The transistor laser is a new laser based on transistor operation. It is not the false but common notion of two forward-biased junctions simply feeding carriers into a common (cavity) region to recombine, but in no sense functioning as a transistor exhibiting β-gain (β = ΔIC/ΔIB) or β compression (i.e., exhibiting base charge-control and the gain connection between input and output of a true transistor).
There are clearly many distinctive features of a transistor laser, in fact—many beyond just the striking gain-compression signature indicating the transition from operation as a three-terminal incoherent spontaneous source to a unique coherent laser light source. The collector junction, besides tilting the carrier population downward from emitter to collector and pinning the base carrier population, reduces the carrier-photon interaction in a TL (the source of Statz-deMars resonance), and thus extends the modulation response speed, yielding a "faster" laser.
The inset in the top portion of the figure on the right shows a scanning electron micrograph picture of the cross-section of a transistor laser (800-µm width) and room-temperature CW operations (center part bullseye). The middle portion of the figure shows the transistor current gain (β = ΔIC/ΔIB, βstim < βspon) vs. the base current and reveals the shift (the "step") in base operation (IB = 45 mA) from spontaneous to stimulated recombination at the threshold of laser operation. The laser spectrum is inserted to show, at IB = 60 mA, the laser operation at λ =1,010 nm. The bottom portion of the figure demonstrates a 19-GHz TL bandwidth and the absence of carrier-photon resonance owing to the fast base carrier lifetime, τB, spon ~ 29 ps.
(Top) Schematic band diagram of a quantum-well (QW) TJ-TL shown with a generic resonator cavity. Note the collector current IC = It + IfkT + IrT consists of the direct transport term It, photon-assisted Franz-Keldysh current IfkT, and regular tunneling IrT. (Bottom) The dependence of optical output of the TJ-TL on VCE indicates the enhancement (VCE < 0.8 V) and quenching (VCE ≥ 0.8 V) of the laser output by FK "self" photon-assisted tunneling (photon absorption). The L-VCE(IB) of the comparison TL (bottom) shows similar behavior occurring gradually except requiring higher bias voltage VCE ≥ 1.6 V.
Tunnel junction transistor laser and signal mixing
The transistor laser acts also as an internal optical-signal detector (internal photon-assisted Franz-Keldysh [FK]effect); the term IfkT in the figure on the right is sensitive to the collector voltage and a mechanism for aiding carrier transfer across the TL base (IE → IC). The top part of the figure shows the schematic band diagram of a tunnel junction transistor laser (TJ–TL) with all the key physical processes labeled. IE is the emitter current (minority carrier injection into the base) with the junction in forward bias; IB is the current resupply of holes by the usual base ohmic contact; IfkT is the internal re-supply of holes by the FK photon-assisted tunneling; IrT represents the resupply of holes via ordinary direct (narrow-junction) tunneling of electrons; and It is the usual base minority carrier current transport of injected electrons that do not recombine and are collected. The collector current is obviously IC = It + IfkT + IrT, with the base hole recombination components adding up as IBr = IB + IfkT + IrT.
In the presence of a stimulated-emission optical field and the laser operation of the TJ-TL, the tunneling process occurs predominantly via Franz-Keldysh (photon-assisted) absorption (IC ≈ It + IfkT). Direct tunneling (not photon-assisted) can be observed at higher VCE biases (IC = It + IfkT + IrT). The collector I-V characteristics of the TJ-TL agree well in form with the optical output, with the L-VCE(IB) characteristics shown in the right-side portion of the bottom figure. In the operation of the TJ-TL under weak collector junction field, collector tunneling (photon-assisted, IfkT > 0) enables the efficient supply of holes to the quantum-well active region, and thus improves the laser optical output by two times that of the comparison TL. We note that the holes supplied by collector tunneling need only relax a distance of roughly 30 nm (from collector to the base quantum-well), as opposed to the lateral distance of 5 mm traversed by holes supplied by an ordinary base ohmic contact (IB). The photon absorption resulting from the weak collector junction field is not sufficient to overcome the photon gain established by emitter and base carrier injection (IE, IB > 0). However, under stronger reverse-biased collector junction field (region 2 of the central portion of the figure), the optical output is reduced and at large bias quenched by Franz-Keldysh absorption.
The collector tunnel junction thus enables the laser output to be controlled effectively by the use of a third terminal control voltage. This is unique to the three-terminal TL and offers signal impedance-matching advantages. Despite relying on only the internal bulk FK effect, the proximity of the collector tunnel junction to the photon generation source (QW) and the strong coupling of the tunneling process to the cavity optical field make possible an effective direct voltage modulation mechanism. This enables the TJ-TL to be directly modulated via a current (δIE, δIB) as well as by voltage control (δVCE, δVBC).
Microwave signal mixing with a common-emitter tunnel junction transistor laser with a pair of input sinusoidal signals: one (f1 = 2.0 GHz) at the base at lower impedance using current modulation, and the other (f2 = 2.1 GHz) at the collector at higher impedance using voltage modulation. The optical output harmonics extend up to 11th order, 4f1+7f2 = 22.7 GHz, despite being limited by amplifier bandwidth.
This is one of the benefits of three-terminal TL operation, multiple input capabilities and better impedance matching. Moreover, the TL collector can be made into an actual tunnel-junction collector (IrT) and provide a sensitive mechanism for fast nonlinear signal mixing and switching. This can be done, for example, over the striking frequency range shown in the figure on the right.
Conclusions and further prospects
Surprising as it is, the transistor laser is still in its infancy—even after a number of papers and patents. Even though there is much more to say and study, in the interest of brevity, we mention merely that the transistor laser promises to become a high-speed-combination coherent-optical signal and coherent-electrical signal integrated circuit element. It has the potential to make possible much faster computers and electronics.
Unlike ordinary transistors, the transistor laser is a three-port device with an electrical input and electrical output, plus it has a coherent optical output. In comparison, the standard transistor is a simpler, more limited two-port element with an electrical input and output. The multi-port capability of the transistor laser makes it necessary in conventional circuit analysis and design to reformulate Kirchhoff's law, taking into account energy conservation and not simply current and charge.
Clearly, the multi-port TL offers much greater topological and device-to-device system design freedom, and it can be foreseen as making possible higher performance combination electrical-optical integrated circuits than can be done with either the transistor itself or even the more limited two-terminal diode.
Just as the point-contact transistor and the transistor idea emerged from Bardeen and Brattain's surface-effect studies and experiments of 1947, the transistor laser came about from the study of the high-current-density high-speed HBT. They both began on one path and emerged on another. Starting and owing its existence to the HBT, the transistor laser, an electrically and optically reinvented quantum-well HBT, now assists in the further understanding and study of the HBT itself. The high-speed HBT, focusing on the critical transistor base (just as did the Bardeen and Brattain point-contact transistor), and taken to the limits of extreme performance, to tiny size and high current density for high-speed operation, has led to a unique three-terminal, three-port transistor laser. The transistor laser, its base region an active high-gain semiconductor medium admitting to ultra-small electrical and (cavity) optical size "compression," may well become a combination ultimate transistor-laser. It is unique as a transistor and as a laser. It is both.
It is not an exaggeration to suggest that the transistor laser, already so revealing and offering still more, may well be the most important thing that has happened to the transistor since the original transistor. It is now 56 years since Bell Labs put Frosch's oxide on silicon and made possible Silicon Valley and today's integrated circuit. It'll be interesting to see in 50 more years what kind of speed and integrated circuits will come from the transistor laser. Now also a laser, the HBT has become much more than just a high-speed transistor; it is also a high-speed three-terminal coherent optical source.
One of us (Milton Feng) is grateful for the support of the Nick Holonyak Jr. Chair of Electrical and Computer Engineering (ECE) and the other (Nick Holonyak Jr.) appreciates the support of the Sony John Bardeen Chair of ECE and Physics. We are also grateful to DARPA and the ARMY-ARO for their research support. We wish to thank the various postdoc and research students listed in the references for their efforts and contributions to this work.
We dedicate this work to John, wherever he is.
Milton Feng and Nick Holonyak Jr. are with the department of electrical and computer engineering at the University of Illinois at Urbana-Champaign, Ill., U.S.A. Holonyak is also affiliated with the department of physics.
References and Resources
>> J. Bardeen and W.H. Brattain. Phys. Rev. 74, 230 (1948).
>> H. Statz and G. deMars, in Quantum Electronics, C.H. Townes, ed., Columbia University Press, N.Y., U.S.A., 1960.
>> E.A. Rezek et al. Appl. Phys. Lett. 31, 534-6 (1977).
>> N. Holonyak Jr. Am. J. Phys. 68, 864-6 (2000).
>> G. Walter et al. Appl. Phys. Lett. 85, 4768-70 (2004).
>> M. Feng et al. Appl. Phys. Lett. 87, 131103 (26 September 2005).
>> M. Feng et al. Appl. Phys. Lett. 88, 063509 (2006).
>> M. Feng and W. Snodgrass. "InP Pseudomorphic Heterojunction Bipolar Transistor (PHBT) with Ft > 750 GHz," IPRM 2007, Japan.
>> M. Feng et al. Appl. Phys. Lett. 95, 033509 (published online 21 July 2009).
>> M. Feng et al. Appl. Phys. Lett. 94, 041118 (published online 29 January 2009).
>> H.W. Then et al. Appl. Phys. Lett. 94, 101114 (published online 13 March 2009).
>> H.W. Then et al. J. Appl. Phys. 107, 094509 (published online 10 May 2010).