letters to nature silicon in such a way that the silicon itself can be used to provide nt it is less than about 10 meV. Consequently, the barrier spatial confinement in three dimensions. To minimize the number the onset of injection is not significantly changed. Any of process steps, in the device described, boron implantation has under steady sta been used both to introduce the dislocation loop array and as the p" direction dynamically replaced by others moving in the forward tve s to form the dislocati species such as the host silicon could be After implantation and annealing, ohmic contacts consisting of and subsequent doping to form the p-n junction can be achieved sample respectively. The ohmic contacts were sintered at 360C for independently. The device operates as a conventional light-emitting 2 min. The top contact was 1 mm in diameter and, as the silicon diode under forward bias ubstrate is still nearly transparent at the emission wavelength of c A simple diagram of the device is shown in the inset to Fig. 1. The 1, 150 nm, a window was left in the larger back contact to allow the 1 X 10cm -at an energy of 30 keV. The sample was subsequently been fabricated Current-voltage(I-V) measurements were made for 20 00C to between the back an to check that the devic form the dislocation loop array and activate the boron dopants. The behaving as a diode. An I-V plot, characteristic of the fully mplants were made into a device grade CZ(Czochralski)n-type processed and working device, is shown in Fig. 1 icon substrate of resistivity 2-40 cm. An important point to The device was mounted into a mechanical holder inside a iphasize at this stage is that all these process steps are entirely continuous flow liquid-nitrogen cryostat. Light from the device conventional and are therefore completely compatible with ULSI was focused into a conventional 0.5 m spectrometer and collected hnology, allowing immediate implementation on a standard by a liquid-nitrogen-cooled germanium p-i-n detector. EL fabrication line. The array of dislocation loops formed has been measurements were then taken from 80 K to above room tempera- observed using cross-sectional transmission electron microscopy. ture. The onset of El was observed as the diode turns on under The array is situated in a planar region parallel to, and around forward bias. No EL is observed under reverse bias. The integrated 100 nm from, the p-n junction. The dislocation loops are typically EL intensity as a function of applied forward voltage is shown in about 80-100 nm in diameter and are spaced around 20 nm apart. Fig. 2 for several temperatures. The turn-on voltage at room The strain field at the edge of the dislocation loop is high and falls off temperature is typical of a conventional silicon diode operating around each loop approxi with the inverse of distance. The under forward injection. The full EL spectra taken, as a function of magnitude and sign of the stress at the edge of the dislocation loop temperature, at a forward current of 50 mA, are shown in Fig 3. The which determines the confining potential can be calculated using low-temperature EL spectrum from the device shows the main the standard elastic theory of dislocations and the known values of features expected from emission at the silicon band edge, including, We obtain a value for the maximum stress at the outside edge (1.2A at 1, 190 nm, the phonon replica of the main peak at 1, 130 nm. The the Poisson's ratio, 0.42, and Youngs modulus, 113 GPa, for silicon room-temperature EL spectrum has the main peak at 1, 160 nm with out)of the interstitial dislocation loop of 25-50 GPa, which will a full-width at half-maximum of 80 nm. No EL or photolumines cause an increase in bandgap energy at the edge of 325-750 mevA cence(PL) is observed at any other wavelengths, even at low omplete description of the stress field across the device, which is temperatures-a result that we again attributed to the close spatial Ist the superposition of the stress field of all the dislocations, is confinement of carriers within regions of bulk silicon. The tempera- complicated and will vary in detail across and between samples. ture dependence of the integrated EL intensity is shown in Fig 4. In However, given the distance dependence of the stress around most systems, demonstrated for light emission in silicon. 4.the EL the loops, the local stress at any point is entirely dominated by the quenches very strongly with increasing temperature, making prac- superposition of the stress field of only the closest dislocations. As tical room-temperature devices problematic. Here, it can be seen carriers are injected across the junction they encounter a blocking that the el intensity actually increases as we go up in temperature. a potential that varies in height depending on the direction of their similar behaviour has been observed but not explained in lase instantaneous velocity, from about 750 meV in the direction of and recrystallized silicon. The device is therefore, as required, most across a loop, down to about 20 me Vat a point equidistant between efficient at room temperature and above. Inset in Fig. 4 is a plot of two adjacent dislocations. It is this blocking potential that confines the integrated PL intensity as a function of measurement tempera- carriers close to the junction region. The varying potential falls away approximately as inverse distance back towards the junction at 80K 9s8m 260K Temperature(K 230 measurement temperature. Inset, a plot of the integrated photoluminescence intensity as Figure 3 Spectra of the electroluminescence intensity against wavelength at various a function of temperature. The solid lines are provided as a guide to the eye. The temperatures. The device was operated at a forward current of 50 ma for all photoluminescence was excited by the 488 nm line of an argon laser at a power of temperatures. NatuReVol4108March2001www.nature.com A@2001 Macmillan Magazines Ltdletters to nature NATURE | VOL 410 | 8 MARCH 2001 | www.nature.com 193 silicon in such a way that the silicon itself can be used to provide spatial confinement in three dimensions. To minimize the number of process steps, in the device described, boron implantation has been used both to introduce the dislocation loop array and as the p￾type dopant to form a p–n junction in an n-type silicon substrate. However, another implant species such as the host silicon could be used to form the dislocations so that the dislocation engineering and subsequent doping to form the p–n junction can be achieved independently. The device operates as a conventional light-emitting diode under forward bias. A simple diagram of the device is shown in the inset to Fig. 1. The device reported here was made by implanting boron at a dose of 1 3 1015 cm 2 2 at an energy of 30 keV. The sample was subsequently annealed in a nitrogen atmosphere for 20 minutes at 1,000 8C to form the dislocation loop array and activate the boron dopants. The implants were made into a device grade CZ (Czochralski) n-type silicon substrate of resistivity 2–4 Q cm. An important point to emphasize at this stage is that all these process steps are entirely conventional and are therefore completely compatible with ULSI technology, allowing immediate implementation on a standard fabrication line. The array of dislocation loops formed has been observed using cross-sectional transmission electron microscopy. The array is situated in a planar region parallel to, and around 100 nm from, the p–n junction. The dislocation loops are typically about 80–100 nm in diameter and are spaced around 20 nm apart. The strain field at the edge of the dislocation loop is high and falls off around each loop approximately with the inverse of distance. The magnitude and sign of the stress at the edge of the dislocation loop which determines the confining potential can be calculated using the standard elastic theory of dislocations10 and the known values of the Poisson’s ratio, 0.42, and Young’s modulus, 113 GPa, for silicon. We obtain a value for the maximum stress at the outside edge (1.2 A˚ out) of the interstitial dislocation loop of 25–50 GPa, which will cause an increase in bandgap energy at the edge of 325–750 meV. A complete description of the stress field across the device, which is just the superposition of the stress field of all the dislocations, is complicated and will vary in detail across and between samples. However, given the inverse distance dependence of the stress around the loops, the local stress at any point is entirely dominated by the superposition of the stress field of only the closest dislocations. As carriers are injected across the junction they encounter a blocking potential that varies in height depending on the direction of their instantaneous velocity, from about 750 meV in the direction of and across a loop, down to about 20 meVat a point equidistant between two adjacent dislocations. It is this blocking potential that confines carriers close to the junction region. The varying potential falls away approximately as inverse distance back towards the junction at which point it is less than about 10 meV. Consequently, the barrier height and the onset of injection is not significantly changed. Any carriers moving back across the junction are, under steady state conditions, dynamically replaced by others moving in the forward direction. After implantation and annealing, ohmic contacts consisting of AuSb eutectic and Al were applied to the back and front of the sample respectively. The ohmic contacts were sintered at 360 8C for 2 min. The top contact was 1 mm in diameter and, as the silicon substrate is still nearly transparent at the emission wavelength of 1,150 nm, a window was left in the larger back contact to allow the electroluminescence (EL) through. Front window devices have also been fabricated. Current–voltage (I–V) measurements were made between the back and front contacts to check that the device is behaving as a diode. An I–V plot, characteristic of the fully processed and working device, is shown in Fig. 1. The device was mounted into a mechanical holder inside a continuous flow liquid-nitrogen cryostat. Light from the device was focused into a conventional 0.5 m spectrometer and collected by a liquid-nitrogen-cooled germanium p–i–n detector. EL measurements were then taken from 80 K to above room tempera￾ture. The onset of EL was observed as the diode turns on under forward bias. No EL is observed under reverse bias. The integrated EL intensity as a function of applied forward voltage is shown in Fig. 2 for several temperatures. The turn-on voltage at room temperature is typical of a conventional silicon diode operating under forward injection. The full EL spectra taken, as a function of temperature, at a forward current of 50 mA, are shown in Fig. 3. The low-temperature EL spectrum from the device shows the main features expected from emission at the silicon band edge, including, at 1,190 nm, the phonon replica of the main peak at 1,130 nm. The room-temperature EL spectrum has the main peak at 1,160 nm with a full-width at half-maximum of 80 nm. No EL or photolumines￾cence (PL) is observed at any other wavelengths, even at low temperatures—a result that we again attributed to the close spatial confinement of carriers within regions of bulk silicon. The tempera￾ture dependence of the integrated EL intensity is shown in Fig. 4. In most systems, demonstrated for light emission in silicon5,6,7 the EL quenches very strongly with increasing temperature, making prac￾tical room-temperature devices problematic. Here, it can be seen that the EL intensity actually increases as we go up in temperature. A similar behaviour has been observed but not explained in laser￾recrystallized silicon9 . The device is therefore, as required, most efficient at room temperature and above. Inset in Fig. 4 is a plot of the integrated PL intensity as a function of measurement tempera- 1,030 1,080 1,130 1,180 1,230 0 4 8 12 16 Wavelength (nm) 80 K 180 K 300 K 110K 260 K Electroluminescence intensity (mV) Figure 3 Spectra of the electroluminescence intensity against wavelength at various temperatures. The device was operated at a forward current of 50 mA for all temperatures. 50 100 150 200 250 300 0 2 4 6 8 10 12 Temperature (K) 50 150 250 4 8 12 Temperature (K) Integrated electroluminescence (a.u.) Integrated photoluminescence (a.u.) Figure 4 A plot of the integrated electroluminescence intensity as a function of measurement temperature. Inset, a plot of the integrated photoluminescence intensity as a function of temperature. The solid lines are provided as a guide to the eye. The photoluminescence was excited by the 488 nm line of an argon laser at a power of 150 mW. © 2001 Macmillan Magazines Ltd