Klr
`
`iiig2
`
`
`
`for the VLSI Era
`Volume 1
`- Process Technology
`Second Edition
`
`ONSEMI EXHIBIT 1008A, Page1
`
`S. Wolf and R.N. Tauber
`
`
`
`ONSEMI EXHIBIT 1008A, Page 1
`
`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`348 SILICON PROCESSING FOR THE VLSI ERA
`
`common dopants in SiO2 can be calculated from measured profiles assuming solutions to Fick's
`second law, with the appropriate boundary conditions.
`The Group III and V elements are known to form glassy networks with SiO2 and accordingly
`their diffusivity strongly depends on their concentration. The diffusivities of these elements are
`very low for concentrations less than 1 %, and generally do not need to be considered in detail.
`A recent paper reports on the diffusion of phosphorus from a phosphorus vapor source into
`thermal oxides. 28 The diffusivity of phosphorus is given by:
`( 2.3)
`-9
`DP= 3.79x10 exp - (cid:173)
`kT
`
`(cm2/sec)
`
`and it was found to be independent of the phosphorus concentration. The solubility of
`phosphorus in the glass, in the temperature range of 1000°C, was found to vary between
`3x 10 20 and 2x 10 22 cm·3• The diffusion mechanism is described as phosphorus dissolving in
`the interstitial sites as P2, where it is incorporated into the network of SiO2. The P2 exchanges
`sites with silicon atoms and continues diffusing through the silicon sites.
`9.6.1 Boron Penetration of Thin Gale Oxides
`In the case of thin gate oxides, the diffusion of boron through the gate oxide must be considered.
`This effect is termed boron penetration. During boron penetration, dopant atoms from the
`heavily-boron-doped polysilicon can diffuse through the thin gate oxide layer and into the
`device channel, where its presence can then change the device threshold. Also, boron
`incorporated in the gate oxide during the diffusion can degrade the oxide breakdown
`characteristics and charge trapping rate. 29 It is further noted that the diffusion of boron through
`oxide is enhanced by the presence of hydrogen and fluorine in the oxide. Fluorine can be
`incorporated into the gate oxide if the boron is introduced into the polysilicon using BF2 as the
`implantation source.
`There are two main ways to reduce boron penetration or diffusion through a thin oxide. The
`most common approach is to incorporate nitrogen into the gate oxide. This technique is known
`to reduce boron diffusion by the nitrogen bonding in the glassy network, which in turn impedes
`the boron flux through the glass. A model has been developed to explain why the presence of
`nitrogen in the oxide hinders the boron diffusion. 30 The model explains that the value of D0
`decreases and that of QA increases as the nitrogen concentration increases.
`The diffusivities (at 900°C) of the common dopants in SiO2 are listed in Chap. 8, Table 8-5.
`There is a class of materials that are fast diffusants in SiO2, including H2, He, OH· 1, Na, 0 2, and
`Ga. Values of D greater than 10-13 cm2/sec have been determined for these elements.
`
`9.7 ANOMALOUS DIFFUSION EFFECTS
`Any deviation from in the diffusion behavior predicted by Fick's Law may be considered as
`anomalous behavior. The most common anomalous diffusion effects include: a) the effect of
`high dopant concentration on diffusion behavior (extrinsic diffusion), as discussed earlier; b) the
`enhancement of diffusion as a result of the built-in electric field; c) the effect of sequential
`diffusions including the "emitter push effect"; d) lateral diffusion under a window; e) oxidation(cid:173)
`enhanced (and retarded) diffusion; and f) transient enhanced diffusion. To a large extent these
`phenomena are dependent upon the interaction of the dopant atoms with silicon point defects.
`Their study helps better understand the interactions between dopants and defects.
`
`ONSEMI EXHIBIT 1008A, Page 58
`
`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`374 SILICON PROCESSING FOR THE VLSI ERA
`
`likelihood of accidents from operating and especially maintaining such equipment, careful
`safety procedures must be established and strictly followed.
`7. Some older-style implanters use diffusion pumps in the beam column, and this will lead to
`organic chemical contamination due to oil backstreaming. The effects of such contamination are
`described in Chaps. 3 and 5.
`10.2 IMPURITY PROFILES OF IMPLANTED IONS
`In order to benefit from the ability to control the number of impurities implanted into a
`substrate, it is necessary to know where the implanted atoms are located after implantation (i.e.,
`it must be possible to predict the depth distribution, or profile of the as-implanted atoms). For
`example, this information is necessary for selecting appropriate doses and energies when
`designing a fabrication process sequence for new or modified integrated circuit devices. What is
`needed to make accurate predictions of implantation profiles is a theoretical model (or models)
`based on the energy interaction mechanisms between the impinging ions and the substrate. In
`this section the topic of how such theoretical models have been developed, and under what
`conditions they provide accurate predictions of implantation profiles, will be addressed. Figure
`10-1 outlines the evolution of the models developed for determining implantation profiles and
`indicates the conditions under which they can be used to provide useful predictions.
`Despite the fact that the derivation of the models is quite mathematical and complex, the
`scope of this text is limited to a more qualitative discussion. Even on a largely qualitative level
`such a presentation is valuable. That is, it provides the reader with an appreciation of the
`intellectual underpinning of ion implantation profile prediction, and also serves as an
`introduction to other physical mechanisms associated with ion implantation. These include
`channeling effects during implantation, substrate damage from implantation, and recoil effects
`that occur when implantations are done through thin layers present on the substrate surface.
`Readers interested in gaining a deeper and more quantitative understanding of implantation
`profile models can refer to references given at the end of the chapter. Some are comprehensive
`surveys, 1,2 while others are papers in which the models were originally published (and which
`thus discuss more fully the assumptions underlying the model, and details of the derivations and
`associated calculations). 3.4,5
`10.2.1 Definitions Associated with Ion Implantation Profiles
`As energetic ions penetrate a solid target material they lose energy due to collisions with atomic
`nuclei and electrons in the target, and eventually come to rest. The total distance that an ion
`travels in the target before coming to rest is termed the range, R. As a result of the collisions
`between the ions and the target material nuclei, this trajectory is not a straight line (and in fact,
`the value of the total distance traveled is not even the quantity that is of highest interest). What
`is of greater interest than R is the projection of this range on the direction parallel with the
`incident beam, since this represents the penetration depth of the implanted ions along the
`implantation direction. This quantity is called the projected range, RP (Fig. 10-2a). As the
`number of collisions and the energy lost per collision experienced by the penetrating ion are
`random variables, ions having the same initial energy and mass will end up spatially distributed
`in the target. An average ion will stop at a depth below the surface given by RP. Some ions,
`however, undergo fewer scattering events than the average, and come to rest more deeply into
`the target. Others suffer more collisions, and come to rest closer to the surface than RP. As a
`
`ONSEMI EXHIBIT 1008A, Page 77
`
`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`390 SILICON PROCESSING FOR THE VLSI ERA
`
`lattice plane (i.e., 60 eV/nm), which is sufficient to displace Si lattice atoms. (Note that although
`electronic stopping is still predominant, nuclear stopping between 10-40 keV can still displace
`Si atoms.) Assuming that one Si atom is displaced per lattice plane for the remainder of the ion
`trajectory, 480 lattice atoms are displaced (i.e., 120 nm/0.25 nm) by the time the boron atom
`comes to complete rest. If each displaced atom is moved roughly 2.5 nm by such collisions, the
`damage volume is found from Vdam = 7t (2.5nm)2(120 nm) = 2.4xl0- 18 cm3 . The damage
`density is 480/V dam= 2xl 020 cm-3, which amounts to only ~0.2% of the atoms. This calculation
`implies that very large doses of light ions are required to produce an amorphous layer, and for
`the most part each ion produces a trail of well-separated primary recoiled Si atoms in the wake
`of the implanted ion. Furthermore, displaced atoms will be separated by short distances from the
`vacancies they leave, because the energies of their recoils arc low. This suggests that only a
`relatively small input of energy to the lattice could cause such separated pairs to rejoin. In fact,
`as will be explained in the following section on annealing, a large fraction of the disorder
`produced during boron implantations is dynamically annealed during the implantation, and
`'" therefore at room temperatures even high-dose boron implantations may not produce an
`amorphous layer. The damage from boron implantations is thus characterized by primary
`crystalline defects. Damage density is distributed versus depth as shown in Fig. 10-16a, which
`shows a sharp buried peak concentration and qualitatively fits the description of the damage(cid:173)
`creation process.
`When heavy ions are implanted, the energy loss is predominantly due to nuclear collisions
`over the entire range of energies experienced by the decelerating heavy ions (Figs. 10-5 and
`10-15). Thus, substantial damage i's expected. Examine the case of 80-kcV arsenic atoms, which
`will have a projected range of ~50 nm. The average energy loss due to nuclear collisions will be
`~1200 cV/nm over the entire range. As a result, the As atoms lose ~300 eV for each Si atomic
`plane that they pass. Most of this energy is transferred to a single lattice atom. The recipient Si
`atom, however, will subsequently produce ~20 displaced lattice atoms. The total number of
`displaced atoms is thus 4000. Again, assuming an average distance moved for each displaced
`atom of ~2.5 nm, the damage volume is Yctam = 7t (2.5 nm) 2(50 nm) = O.Sxl0- 18 cm3. The
`damage density is then 4000/V dam = 5 x 1021 cm-3, or ~ I 0% of the number of atoms in the
`lattice within the damage volume. Since a single ion is capable of producing such heavy
`damage, it is reasonable to expect that some local regions of a silicon substrate (which when
`subjected to even light doses of heavy-ion bombardment) will suffer enough damage to become
`amorphous. Some damage density distributions due to heavy-ion (e.g., As) implants are shown
`in Fig. IO- I 6b. They exhibit a broad buried peak that is a replica of recoiled range distribution.
`10.3.3 Amorphous Layer Damage
`Simple qualitative concepts illustrate how continued bombardment by heavy ions will lead to
`the formation of continuous amorphous layers. That is, heavy-ion damage accumulates with ion
`dose through an increase in the density of localized amorphous regions. Eventually these
`regions overlap, and a continuous amorphous layer is the result. The evolution of a continuous
`amorphous layer from the accumulation and overlap of damage formed by individual atoms has
`been observed (Fig. 10-17) using Rutherford backscattering spectroscopy in a channeling mode.
`In Fig. 10-l7a it can be seen that damage produced by 1.7 MeV Ar+ ions in Si builds up to an
`initial damage distribution with a peak at a depth of ~ 1.3 µ m. At that depth, individual
`amorphous zones are likely to be created by each ion (Fig. 10-17b). Closer to the surface the
`damage consists predominantly of isolated defect clusters (akin to the damage caused by light
`
`ONSEMI EXHIBIT 1008A, Page 93
`
`

`

`

`

`

`

`

`

`394 SILICON PROCESSING FOR THE VLSI ERA
`
`where q is the electronic charge. Although µ depends strongly on the concentration of doping
`atoms and implantation damage, values of Rs have been tabulated utilizing known mobility
`data. 31 For a known dose, full electrical activity is reached when the predicted Rs is reached.
`Electrical activation of implanted impurities in amorphous layers proceeds differently than in
`layers with primary crystalline damage. As will be discussed, electrical activation in amorphous
`layers occurs as the impurities in the layer are incorporated onto lattice sites during recrys(cid:173)
`tallization. Electrical activation in crystalline damaged regions exhibits more complex behavior.
`For example, Fig. 10-20 shows the isochronal electrical activation behavior of implanted
`boron (i.e., anneals performed at varying temperatures, but for identical times). In this curve, the
`measured surface carrier concentration (normalized for different junction depths to the dose, in
`cm-2) is used to indicate the degree of activation. That is, when P1-1aiil<P = 1, full activation is
`reached. Note that other impurities exhibit similar behavior to that shown in Fig. 10-20,
`provided the implantation does not cause a continuous amorphous layer to be formed.
`The temperature range up to 500°C (Region 1 of Fig. 10-20) shows a monotonic increase in
`~ electrical activity. This is due to the removal of trapping defects and the concomitant large
`increase in free carrier concentration as the traps release the carriers to the valence or
`conduction bands. In Region 2 (500-600°C), substitutional B concentration actually decreases.
`This is postulated to occur as a result of the formation of dislocations at these temperatures.
`Some boron atoms that were already on substitutional sites are believed to precipitate on or near
`these dislocations. In Region 3 (>600°C), the electrical activity increases until full activation is
`achieved at temperatures -800 1000°C. The higher the dose, the more disorder, and the higher
`the final temperature required for full activation. At such elevated temperatures, Si self(cid:173)
`vacancies are generated. They migrate to the B precipitates, allowing boron to dissociate and fill
`the vacancy (i.e., a substitutional site).
`Activation of implanted impurities by rapid thermal processing (RTP, see Chap. 9) has also
`been studied. The time-temperature cycle to reach minimum sheet resistance for As, P, and B is
`~5-10 sec at 1000-1200°C, the exact condition being dependent on implanted species, energy,
`and dose. 33 ,34
`10.3.4.2 Annealing of Primary Crystalline Damage: Isolated point defects and point defect
`clusters (that predominantly occur during light ion implantation), and locally amorphous zones
`(that are typically observed from light doses of heavy ions), are both regions of primary
`crystalline damage that exhibit comparable annealing behavior. At low temperatures (up to
`-500°C}, vacancies and self-interstitials that are in close proximity undergo recombination,
`thereby removing trapping defects. At higher temperatures (500-600°C), as described above,
`dislocations start to form, and these can capture impurity atoms. Temperatures of 900-1000°C
`are required to dissolve these dislocations. Note that the activation energy of impurity diffusion
`in Si is always smaller than that of Si self-diffusion (see Chap. 9). Therefore, the ratio of defect
`annihilation to the rate of impurity diffusion becomes greater as the temperature is raised. This
`implies that the higher the anneal temperature the better, with the upper limit being constrained
`by the maximum allowable junction depth dictated by the device design. 30
`It is also important that the steps used to anneal implantation damage be conducted in a neu(cid:173)
`tral ambient, such as Ar or N2. That is, dislocations which form during annealing35 can serve as
`nucleation sites for oxidation induced stacking faults (OISF, see Chap. 2) if oxidation is carried
`out simultaneously with the anneal (i.e., the annealing is performed in an oxygen ambient).
`
`ONSEMI EXHIBIT 1008A, Page 97
`
`

`

`

`

`396 SILICON PROCESSING FOR THE VLSI ERA
`
`Fig, 10·21 Solid phase regrowth of a 200 keV, 6xl0 15tcm2 antimony implantation at 525°C. TEM cross
`section micrograph. Courtesy of Institute of Physics, Conference Series. 36
`layer. The gives rise to optical interference effects from the light reflected off the subsurface
`damage layers and the surface.
`Some of the crystalline defects in the region beyond the amorphous layer are annealed out
`during subsequent thermal cycles, but others give rise to extended defects (such as dislocation
`loops and stacking faults), which then grow and interact. Under some implantation and
`annealing conditions, these defects move to the surface and eventually disappear, while in others
`they grow into larger structures which intersect the surface or remain in the bulk.
`In a large number of cases, the total number of "excess" Si atoms found in dislocation loops
`after thermal annealing fits a remarkably simple model. Even though hundreds to thousands of
`atoms are displaced from their positions in the Si lattice by each ion impact, the number of
`residual Si atoms left out of the lattice after the completion of thermal annealing increases with
`ion dose and is closely proportional to the number of ions. This is known as the "+l" model,
`where the number of excess Si interstitial atoms is equal to the number of implanted dopant
`atoms that occupy lattice sites after thermal annealing. 38 ,39 These Si interstitials arise from the
`dopant atom replacing the Si-atom in the lattice.
`10.3.f4.f4 Dynamic Annealing Effects: The heating of the wafer during implantation can impact
`the implantation damage and the effects of subsequent annealing. A rise in temperature
`increases the mobility of the point defects caused by the damage, and this gives rise to healing
`of damage even as the implant process is occurring, hence the name dynamic annealing. 40 In the
`case of light ions, sufficient damage healing may occur to prevent the formation of amorphous
`layers, even at very high implantation doses. In the case of heavy ion implantations, dynamic
`annealing can cause amorphous layer regrowth during the implantation step.
`A study by Prussin et al., 41 showed that the wafer cooling capability of an ion implanter can
`impact the structure of the damage following implantation because of dynamic annealing
`effects. That is, if a wafer is prevented from being significantly heated above room temperature
`by adequate heat sinking during implantation, dynamic annealing is minimized. On the other
`hand, if no heat sinking is provided and wafers are allowed to rise to temperatures ~150-300°C,
`dynamic annealing effects can produce changes in implantation damage structures. This
`typically occurs in non-reproducible and unwanted ways, such as the formation of buried
`amorphous layers, or crystalline layers containing high densities of dislocation loops.
`10.3.f4.5 Diffusion of Implanted Impurities: As described in Chap. 9, the diffusion of impurities
`in single-crystal Si is a complex phenomenon. The diffusion of impurities in implanted Si is
`
`ONSEMI EXHIBIT 1008A, Page 99
`
`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`

`