You are on page 1of 7

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 59, NO.

1, JANUARY 2012 87

Temperature Compensation of Silicon Resonators via


Degenerate Doping
Ashwin K. Samarao, Member, IEEE, and Farrokh Ayazi, Senior Member, IEEE

Abstract—This paper demonstrates the dependence of temper- con microresonator-based oscillator in a feedback loop and to
ature coefficient of frequency (TCF) of silicon micromechanical tune its frequency electronically to compensate for temperature
resonators on charge carrier concentration. TCF compensation is changes [7]. This is somewhat more challenging to accomplish
demonstrated by degenerate doping of silicon bulk acoustic res-
onators (SiBARs) using both boron and aluminum dopants. The in bulk- than in flexural-mode resonators because the electrosta-
native TCF of −33 ppm/◦ C for silicon resistivity of > 103 Ω · cm tic tunability of the bulk-mode resonators is substantially less
is shown to reduce to −1.5 ppm/◦ C at ultralow resistivity of than that of the flexural-mode resonators [8]. Another possible
∼ 10−4 Ω · cm using relatively slow diffusion-based boron dop- approach is to enclose the resonator in a thermally isolated
ing. However, the faster thermomigration-based aluminum doping “micro oven” whose temperature is kept constant through the
offers TCF reduction to as low as −2.7 ppm/◦ C with much use of heating elements [9], [10] or by running a current
reduced processing time. A very high Q of 28 000 at 100 MHz is
measured for a temperature-compensated SiBAR. through the structural body of the resonator [8]. However,
need for additional circuitry (and thereby chip area) and an
Index Terms—Aluminum thermomigration, degenerate increase in the overall power consumption are the two most
boron doping, silicon micromechanical resonators, temperature
compensation.
significant problems in these active temperature compensation
techniques. Hence, it is desirable to have passive techniques
I. I NTRODUCTION that compensate for most of the TCF, if not entirely, so that
active compensation techniques can be used in conjunction to
A FTER nearly two decades of research and development,
the manufacturability and reliability of silicon resonator
technology are successfully established to enable their wide-
compensate for the rest to achieve zero-TCF resonators without
excessive power consumption and calibration.
The most common passive TCF compensation technique is
spread commercialization and insertion. Currently, they are
based on using a composite structure, such as silicon dioxide
being used as drop in replacements for quartz crystals as well
and silicon, whose respective stiffness changes with temper-
as integrated modules for frequency and timing applications [1],
ature in opposite ways. Such an approach of using a positive
[2]. Their small size, ease of fabrication, frequency scalability,
TCF material to compensate for negative TCF dates back to the
and high-quality factors (≥ 10 000) along with moderate mo-
early 1980s [11] but is still the object of much active research
tional resistances (∼ a few 100 Ω at ≥ 100 MHz) persuaded
[12]–[14]. However, the acoustic loss at the interface of the
efforts to tailor these devices for oscillators and resonant sensor
different materials in the composite structure loads quality
applications [3]–[5]. However, one of the most dominant draw-
factor Q, while the stress mismatch at such an interface might
backs of silicon resonators has been their large temperature
lead to hysteresis. Further, dielectric charging has been reported
coefficient of frequency (TCF). The resonance frequency of
in thermally oxidized capacitive silicon resonators, which are
the silicon microresonator decreases with increasing temper-
known to cause a drift in frequency over time [15]. Thus, since
ature, thereby exhibiting a negative yet almost linear TCF of
passive temperature compensation techniques are desired, an
approximately −30 ppm/◦ C. Such a TCF is significantly larger
alternative to composite resonator structures is needed. This
in magnitude than that of the worst AT-cut quartz crystals [6].
paper addresses the issue by identifying the dependence of
Hence, considerable research has been carried out to reduce the
TCF on charge carrier concentration in the resonating silicon
TCF of single-crystal silicon (SCS) microresonators.
microstructure and proposes degenerate doping of silicon for
There exist system-level approaches to temperature com-
TCF compensation.
pensation that require tuning or ovenization of the resonator.
One approach to temperature compensation is to use the sili-
II. SiBAR
Temperature compensation via degenerate doping is demon-
Manuscript received July 11, 2011; revised October 7, 2011; accepted strated using single-crystal silicon bulk acoustic resonators
October 11, 2011. Date of publication November 3, 2011; date of cur-
rent version December 23, 2011. This work was supported by Integrated (SiBARs) (see Fig. 1). A SiBAR consists of a resonating sus-
Device Technology (IDT), Inc. The review of this paper was arranged by pended SCS bar separated by narrow air gaps from two identical
Editor A. M. Ionescu. drive/sense polysilicon electrodes. The suspended microstruc-
A. K. Samarao is with Robert Bosch Research and Technology Center, Palo
Alto, CA 94304 USA. ture is supported using narrow tethers at its shorter edges,
F. Ayazi is with the School of Electrical and Computer Engineering, Georgia which also provide electrical contact to dc polarization voltage
Institute of Technology, Atlanta, GA 30332-0250 USA. Vp that is applied at the Vp pads. Now an ac signal applied
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. at the drive electrode results in a time-varying electrostatic
Digital Object Identifier 10.1109/TED.2011.2172613 force applied across the air gap onto the corresponding face

0018-9383/$26.00 © 2011 IEEE


88 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 59, NO. 1, JANUARY 2012

Fig. 1. SEM image of a 100-MHz SiBAR. (W = 40 μm; L = 10 × W =


400 μm; T = 10 μm). E and ρ are the Young’s modulus and density of silicon.
Fig. 3. Illustration showing (a) equivalent HH and LH energy surfaces of
the valence band in silicon that are (b) permanently split due to the large
compressional strain from the very high boron doping of the silicon crystal.
E and k are the electronic energy and wave vector, respectively.

favorable light-hole (LH) band [see Fig. 2(b)] [17], [18]. The
net flow of holes and thereby the electronic energy of the system
increases with increasing temperature. A similar analogy exists
for the electrons in the conduction band of silicon. Per the
conservation of energy principle, such a temperature-dependent
increase in the electronic energy manifests itself as a corre-
sponding decrease in the elastic energy of the system. Thus,
a progressive reduction in the stiffness (i.e., Young’s modulus
(E)) of the resonating silicon microstructure and thereby its
resonance frequency is observed with increasing temperature
(i.e., TCF). Although linear thermal expansion coefficient α of
Fig. 2. Illustration showing (a) equivalent HH and LH energy surfaces of silicon also contributes to the TCF, its contribution is negligible
the valence band in silicon that contain most of the holes at steady-state and
(b) propagation of acoustic waves splits the equivalent surfaces leading to a
compared with the temperature coefficient of Young’s modulus
flow of holes from HH to LH. E and k are the electronic energy and wave (TCE). Temperature compensation techniques that have been
vector, respectively. reported to date combat the effect of TCE, whereas this work
targets the manipulation of charge carriers that cause it in the
of the resonator. At target frequency f0 determined by W , first place [19].
the resulting width-extensional mode (WEM) of resonance [see Thus, a momentary strain produced by the propagation of the
Fig. 2(b)] modulates the transduction air gap on the other side acoustic waves is understood to create a temperature-dependent
inducing a voltage on the sense electrode. The SiBARs used in change in the electronic and elastic energies of the system. Such
this work were transduced using very narrow 100-nm air gaps an effect of the momentary strain could be made minimal in
fabricated using the HARPSS process [16]. comparison if a larger permanent strain could be created in the
silicon crystal. For example, a boron atom that has a smaller
radius than silicon atom strongly bonds to only three of the four
III. TCF C OMPENSATION VIA D EGENERATE D OPING
adjacent silicon atoms when diffused into the crystal lattice. At
The propagation of an acoustic wave through a solid material very high doping levels, such an atomic arrangement produces
is typically characterized by an alternating set of compressional a strong shear strain in the silicon lattice, which is sufficient to
and dilational forces that perturb the periodicity of the atomic create a large permanent separation between the equivalent LH
lattice during wave propagation. In a semiconductor such as sil- and HH valence bands. As a result, the majority of the holes
icon, such a perturbation of atomic periodicity directly impacts permanently shift from the HH to the LH band, thereby almost
its electronic band structure. For example, at steady-state, the depleting the holes from the former (see Fig. 3) [20].
valence band of silicon is made up of three energy surfaces The effect of any additional band splitting caused by the
in k-space, two of which are equivalent and are energetically propagation of acoustic waves in such a highly-doped SiBAR
favorable to contain almost all the holes [see Fig. 2(a)]. The now becomes minimal in comparison for two reasons: first, very
resulting longitudinal strain due to the WEM of resonance in few holes are available in the almost depleted HH band of a
a SiBAR splits the equivalent energy bands, leading to a flow highly doped silicon to flow to the LH band; second, the energy
of holes from the heavy-hole (HH) to the more energetically required to flow from HH to LH is much larger compared with a
SAMARAO AND AYAZI: TEMPERATURE COMPENSATION OF SILICON VIA DEGENERATE DOPING 89

TABLE I
M EASURED TCF IN SiBARs FABRICATED ON S ILICON WAFERS W ITH
V ERY L OW TO V ERY H IGH B ORON D OPING L EVELS

Fig. 4. Illustration of the minimized flow of holes in the valence band of


silicon during acoustic transduction in a very highly boron-doped SiBAR.

nonhighly doped silicon, thereby making such a transition least


energetically favored (see Fig. 4).
As a result, the momentary flow of holes from HH to LH is
significantly minimized. Thus, a considerable reduction in the
variation of the electrical and elastic energies of the system with
temperature is effected, thereby resulting in TCF compensation.
At degenerate levels of boron doping, the acoustic wave can
potentially be shielded entirely from the k-space contours of the
valence band, which, in turn, should completely compensate the
TABLE II
TCE component of the TCF, i.e., B ORON D OPING R ECIPE U SING S OLID AND L IQUID B ORON S OURCES
TCE + α
TCF = . (1)
2
The native silicon TCE of ∼ −57 ppm/◦ C and α of ∼
−3 ppm/◦ C yield a TCF of ∼ −30 ppm/◦ C in silicon microres-
onators. Upon complete TCE compensation via degenerate
doping, a TCF as low as ∼ −1.5 ppm/◦ C could be achieved.

IV. TCF C OMPENSATION VIA D EGENERATE


by repeated doping/annealing processes. For example, a boron-
B ORON D OPING
doped resistivity of ∼ 0.01 Ω · cm resistivity has been mea-
Both Czochralski and float-zone silicon wafer manufacturing sured to reduce to ∼ 0.001 Ω · cm after five repetitions of the
processes typically involve an in situ doping with either n- or solid boron dope/anneal recipe shown in Table II. For further
p-type dopants. Such predoped silicon wafers can be procured reduction in resistivity, liquid spin-on-dopant boron sources
as is from the wafer vendors to verify the TCF compensation were used [Futurrex Inc. (USA), BDC1-2000]. After six repeti-
in silicon microresonators with increasing doping levels (i.e., tions of spin-on-dope/anneal recipe, the boron-doped resistivity
decreasing resistivities). Thus, the proposed solution of very could be further reduced from ∼ 0.001 to ∼ 0.0001 Ω · cm. In
high doping for TCF compensation can be easily verified for both cases of solid and liquid dopants, the borosilicate glass
any SCS microresonator design with no additional need for layer that forms after every dope/anneal cycle was removed us-
processing capabilities to dope silicon. ing hydrofluoric acid. All the reported resistivities in this paper
Table I summarizes the measured TCF across various resis- were accurately measured using four-point probe measurement.
tivities from SiBARs fabricated on a 20-μm-thick boron-doped Although longer hours of processing are needed to achieve a
device layer of SOI wafers as-procured from Ultrasil Corp. resistivity of ∼ 0.0001 Ω · cm (see Table II), such a resistivity
(USA). Although only a minimal TCF reduction of ∼ 4 ppm/◦ C is lower than any commercially available silicon wafer. At such
is observed from very high to low resistivity silicon substrates, doping levels, the diffusion depth of boron in silicon is typically
substantial TCF reduction has been measured at very low and limited to 7 ∼ 8 μm [22]. Hence, 5-μm-thick substrates were
ultralow resistivities. As shown in Table I, an overall TCF chosen to fabricate uniformly doped SiBARs at such degenerate
reduction of 24 ppm/◦ C can be readily achieved in silicon doping levels of boron in silicon. A 5-μm-thick 0.0001 Ω · cm
microresonators by opting for ultralow resistive substrates. No SiBAR is found to measure a TCF of −3.56 ppm/◦ C, which is
significant impact was measured on the Q and insertion loss of an order of magnitude lower than that measured at a very high
these microresonators across such wide range of doping levels. resistivity of 1000 Ω · cm (see Fig. 5). The increase in reso-
It is worth mentioning that such very low and ultralow nance frequency after heavy doping is due to the increase in the
resistivity substrates were specially ordered for the purpose Young’s modulus of silicon and is indicative of the degenerative
and are typically more expensive compared with the relatively doping [23]. As seen from the square of correlation coefficient
higher resistivity substrates. Such very high levels of doping of linearity R2 reported in Fig. 5, the linearity of the TCF curve
can be also achieved starting from a low resistivity substrate is compromised to some extent due to degenerate boron doping.
90 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 59, NO. 1, JANUARY 2012

Fig. 5. Order of magnitude reduction in TCF measured at a boron-doped


SiBAR resistivity of 0.0001 Ω · cm compared with 1000 Ω · cm.

Fig. 7. SEM images showing the observed outdiffusion of boron from the
silicon surface after seven repetitions of the spin-on-dope/anneal processes.

Fig. 8. Saturation of TCF at ∼ −1.5 ppm/◦ C of the 5-μm-thick SiBAR after


seven repetitions of spin-on-dope/anneal processes.
Fig. 6. Measured response of the degenerately doped SiBAR with a TCF of
−3.56 ppm/◦ C showing a quality factor Q of 33 000 in vacuum. complete compensation of TCE [see (1)]. It is to be noted that
such outdiffusion of boron results in lattice damage to silicon
Such degenerately doped 0.0001 Ω · cm 5-μm-thick SiBARs that reflects as a relatively reduced Q of 20 000 in vacuum.
measure a Q of 33 000 (versus 35 000 for a nondegenerately Thus, degenerate boron doping has been successfully demon-
doped 5-μm-thick SiBAR with a resistivity of 0.01 Ω · cm) at strated to offer a reduction from a native TCF of −33 to
Vp of 20 V in vacuum (see Fig. 6). This indicates a minimal −1.5 ppm/◦ C in silicon micromechanical resonators. Similar
effect on the acoustic loss in silicon microresonators from temperature compensation has been also recently demonstrated
longer hours of repeated thermal cycling during the doping in thermally actuated silicon microresonators using degener-
process. ate phosphorous doping [25]. However, the longer hours of
However, upon any further doping, boron is observed to processing needed to achieve such degenerate doping levels call
outdiffuse from the surface of the silicon, leaving behind a for a faster alternative. Further, the difficulty in achieving very
relatively porous silicon surface. Such a surface of silicon high doping levels for thicker substrates limits the proposed
as-observed after seven repetitions of the spin-on-dope/anneal TCF compensation technique to substrates that are thinner than
processes is shown in Fig. 7. The outdiffusion is indicative of 7 ∼ 8 μm. The next section addresses both these issues by
doping levels that surpass the solid solubility limit of boron presenting a boron-assisted aluminum doping technique that
in silicon (∼ 2 × 1020 atoms/cm3 ). At such levels of doping, is capable of achieving very high doping levels within few
the resistivity values could not be accurately measured and minutes of processing in much thicker substrates.
hence are not reported. The SiBARs fabricated at such doping
levels measure the lowest recorded TCF of −1.43 ppm/◦ C.
V. TCF C OMPENSATION VIA B ORON -A SSISTED
Interestingly, no further reduction in TCF was measured even
A LUMINUM D OPING
after ten repetitions of spin-on-dope/anneal processes (see
Fig. 8). Interestingly, Wang et al. had theoretically predicted Boron diffuses as an interstitial dopant [26], which demands
a positive TCF with increasing levels of boron doping in silicon for longer hours of annealing to become electrically active. On
microresonators [24]. To the contrary, this work experimentally the other hand, aluminum (also a p-type dopant) becomes read-
proves that positive TCFs were not measured even at degenerate ily electrically active by diffusion via self-interstitial mecha-
levels of boron doping in silicon. Further, the lowest measured nism [27]. It has been shown that aluminum can thermomigrate
TCF value of ∼ −1.5 ppm/◦ C is understood to be the sole against a temperature gradient into hundreds-of-micrometers-
contribution of linear thermal expansion component α after thick silicon within few tens of minutes [28]. Thermomigration
SAMARAO AND AYAZI: TEMPERATURE COMPENSATION OF SILICON VIA DEGENERATE DOPING 91

Fig. 9. Schematic of the wafer-level aluminum thermomigration process.

is typically performed by patterning aluminum on one side of


the wafer while the other side is radiatively heated to approxi-
mately 1000 ◦ C. Such one-sided heating results in a temperature
gradient across the thickness of the wafer. Once above the
eutectic point (577 ◦ C), the aluminum melts and dissolves some
of the silicon at the surface. The resulting droplet follows the
thermal gradient toward the hot surface. As the droplet moves,
Fig. 10. Schematic of the device-level aluminum thermomigration process.
the warmer side of the droplet dissolves more silicon, and the (a) Released SiBAR. (b) Blanket deposition of 500 Å aluminum. (c) Joule
cooler side of the droplet epitaxially resolidifies; the resolidified heat the support tethers and SiBAR at 120 mA for 10 min. (d) Aluminum
material has an aluminum concentration determined by the thermomigrated SiBAR at room temperature; wirebond traces are visible in
(d) since the same device was used for SEM images.
solid solubility limit for aluminum in silicon at the solidification
temperature, which is approximately 2 × 1019 atoms/cm3 [29].
The process is known to be highly anisotropic and relatively aluminum thermomigration was carried out postfabrication.
fast, with droplets traveling through a 300-μm-thick silicon Hence, stress due to thermal gradients impacts the Q and
wafer in less than 10 min [30]. The uniformity and speed of long-term stability of these resonators. On the other hand,
aluminum thermomigration is known to be enhanced by the a device-level thermomigration process could circumvent the
presence of boron atoms in silicon [27]. Further, both boron problem while offering the expected reduction in TCF. Fig. 10
and aluminum dopants are electrically active and will contribute shows the schematic and SEM images of such a device-level
to TCE compensation. Thus, such a boron-assisted aluminum aluminum thermomigration process developed for SiBARs. A
thermomigration is a faster alternative to degenerate boron 500-Å-thin layer of aluminum was evaporated onto the SiBAR
doping for TCF reduction. [see Fig. 10(b)]. A current of 120 mA was passed for 10 min
This was investigated by evaporating 200 Å of aluminum through the SiBAR resonator element via the support tethers
onto the surface of 20-μm-thick SiBARs fabricated on a boron- [see Fig. 10(c)] [31]. At such high currents, the narrow support
doped ∼ 0.01 Ω · cm silicon wafer. The required temperature elements Joule heat to ∼ 1000 ◦ C while the wider SiBAR
gradient is created in a furnace by turning off one of the reaches the Al–Si eutectic temperature, thereby creating the
many coil heaters of the furnace (see Fig. 9). The wafer is necessary temperature gradient [see Fig. 10(c)]. Aluminum
positioned in the furnace at the boundary of the hot and cold on top of the relatively cold SiBAR laterally diffuses through
regions, with the aluminum-deposited side facing away from the silicon toward the hot support elements, thereby doping it
the heat. The temperature gradient across the thickness of the [see Fig. 10(d)]. A closer look at the thermomigrated surface
wafer drives the aluminum to thermomigrate into the resonating (see Fig. 11) shows regions of Al–Si eutectic alloy alongside
silicon bulk. After 1 h of thermomigration in nitrogen ambient, exposed silicon where aluminum has completely diffused into
the TCF of the SiBAR reduces by 24 ppm/◦ C, from −27.8 to the bulk of the resonator. The short duration of the process is
−3.8 ppm/◦ C. insufficient to thermomigrate the entire thickness of aluminum,
The reduced TCF is equivalent to that measured at a boron- thereby leaving behind some residues (due to thicker than
doped resistivity of ∼ 0.0001 Ω · cm (see Fig. 6). Thus, a boron desired Al), which could be thermomigrated with further Joule
doping duration of approximately 90 h (see Table II) for a heating. Although aluminum is deposited on the Vp pads as well
similar TCF reduction on a 5-μm-thick substrates can be alter- [see Fig. 10(b)], its relatively larger cross section remains below
natively realized using an hour of aluminum thermomigration the Al–Si eutectic temperature during Joule heating. Hence, no
on substrates that are 20 μm or thicker. On the downside, thermomigration has been observed to occur on the Vp pads.
due to the rapid thermomigration and the large temperature A reduction in TCF by 14 ppm/◦ C (from −22.13 to
gradient across the entire wafer, some structural damage could −7.93 ppm/◦ C) has been measured over 10 min of device-level
be observed on the surface of the resonators. As a result, a thermomigration on a ∼ 0.001 Ω · cm SiBAR. From the values
considerably reduced Q of 16 000 has been observed at vacuum in Table I, it can be inferred that a TCF value corresponding
in wafer-level thermomigrated SiBARs. to very low resistivity of a silicon wafer can be reduced to its
Unlike boron doping that is performed on a plain SOI ultralow resistive counterpart within few minutes of aluminum
wafer prior to the fabrication of the resonators, the wafer-level thermomigration. The passage of current for longer durations
92 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 59, NO. 1, JANUARY 2012

Fig. 13. TCF of a 20-μm-thick [∼ 0.0001 Ω · cm; boron-doped] SiBAR


before and after device-level thermomigration with 500 Å of aluminum at
120 mA for 10 min.
Fig. 11. SEM showing a close-up of the aluminum thermomigrated SiBAR in
Fig. 10(d).

Fig. 14. Measured response of a 20-μm-thick [∼ 0.0001 Ω · cm; boron-


doped] SiBAR after device-level thermomigration with 500 Å of aluminum at
120 mA for 10 min.
Fig. 12. Measured response of a 20-μm-thick [∼ 0.001 Ω · cm; boron-doped]
SiBAR after device-level thermomigration with 500 Å of aluminum at 120 mA
for 10 min. VI. C ONCLUSION
might lead to plastic deformation at the narrow support tethers, Temperature compensation of silicon micromechanical res-
thereby causing a failure of resonator operation. However, if onators has been achieved using degenerate boron doping
the supports are redesigned to be wider, thermomigration could and boron-assisted aluminum doping. A starting TCF of
be performed for longer durations to achieve further reduction −33 ppm/◦ C is reduced to −1.5 ppm/◦ C by degenerate boron
in TCF. Unlike boron doping and wafer-level thermomigra- doping of the silicon microresonator. Both wafer- and device-
tion processes, device-level aluminum thermomigration is best level aluminum thermomigration techniques are developed as
suited to be a postfabrication postpackaging technique, wherein faster alternatives to the relatively slower diffusion-based boron
the TCF can be programmed to a desired value using Joule doping process. Although the former is found to compromise
heating as an electrical calibration step. on the Q due to stress from the thermal gradient, the latter
Minimal or no structural damage has been observed from the circumvents the problem and offers low TCF along with low
device-level thermomigration process as evident from a Q of insertion loss and high Q.
29 000 measured from these devices in vacuum. The measured
insertion loss of 39.5 dB is typical of 20-μm-thick SiBARs with
a capacitive air gap of ∼100 nm at Vp of 20 V (see Fig. 12). ACKNOWLEDGMENT
A similar device-level thermomigration was also performed The authors would like to thank the staff of the Nanotech-
on a 20-μm-thick degenerately boron-doped ∼ 0.0001 Ω · cm nology Research Center (NRC) at the Georgia Institute of
SiBAR. The TCF curve before thermomigration exhibits an Technology for their assistance.
interesting anomalous behavior, as shown in Fig. 13. This is
thought to arise due to the nonuniform doping profile at such
very high doping levels along the 20-μm thickness of the de- R EFERENCES
vice. Aluminum thermomigration is found to restore the linear- [1] F. Ayazi, “MEMS for integrated timing and spectral processing,” in Proc.
IEEE CICC, 2009, pp. 65–72.
ity of the TCF curve and offer a TCF as low as −2.72 ppm/◦ C [2] B. Kim, R. Melamud, R. A. Candler, M. A. Hopcroft, C. M. Jha,
while maintaining a high Q of 28 000 in vacuum (see Fig. 14). S. Chandorkar, and T. W. Kenny, “Encapsulated MEMS resonators—A
SAMARAO AND AYAZI: TEMPERATURE COMPENSATION OF SILICON VIA DEGENERATE DOPING 93

technology path for MEMS into frequency control applications,” in Proc. [25] A. Hajjam, A. Rahafrooz, and S. Pourkamali, “Sub-100 ppm/◦ C temper-
IEEE IFCS, 2010, pp. 1–4. ature stability in thermally actuated high frequency silicon resonators via
[3] H. M. Lavasani, A. K. Samarao, G. Casinovi, and F. Ayazi, “A 145 MHz degenerate phosphorous doping and bias current optimization,” in IEDM
low phase-noise capacitive silicon micromechanical oscillator,” in IEDM Tech. Dig., 2010, pp. 170–173.
Tech. Dig., 2008, pp. 675–678. [26] O. Krause, H. Ryssel, and P. Pichler, “Determination of aluminum diffu-
[4] R. Tabrizian, M. Rais-Zadeh, and F. Ayazi, “Effect of phonon interactions sion parameters in silicon,” J. Appl. Phys., vol. 91, no. 9, pp. 5645–5649,
on limiting the f Q Product of micromechanical resonators,” in Proc. May 2002.
Transducers, 2009, pp. 2131–2134. [27] B. Sadigh, T. J. Lenosky, S. K. Theiss, M.-J. Caturla, T. Diaz de La Rubia,
[5] B. P. Harrington, R. Abdolvand, A. Hajjam, J. C. Wilson, and and M. A. Foad, “Mechanism of boron diffusion in silicon: An ab initio
S. Pourkamali, “Thin-film piezoelectric-on-silicon particle mass sensor,” and kinetic Monte Carlo study,” Phys. Rev. Lett., vol. 83, no. 21, pp. 4341–
in Proc. IEEE IFCS, 2010, pp. 238–241. 4344, Nov. 1999.
[6] W.-T. Hsu and C. T.-C. Nguyen, “Stiffness-compensated temperature- [28] W. Pfann, Zone Melting. Huntington, NY: Krieger, 1978.
insensitive micromechanical resonators,” in Proc. IEEE Int. Conf. MEMS, [29] D. Lischner, H. Basseches, and F. D’Altroy, “Observations of the tem-
2002, pp. 731–734. perature gradient zone melting process for isolating small devices,” J.
[7] J. C. Salvia, R. Melamud, S. A. Chandorkar, S. F. Lord, and T. W. Kenny, Electrochem. Soc., vol. 132, no. 12, pp. 2997–3001, Dec. 1985.
“Real-time temperature compensation of MEMS oscillators using an inte- [30] C. C. Chung and M. G. Allen, “Thermomigration-based junction isolation
grated micro-oven and a phase-locked loop,” J. Microelectromech. Syst., of bulk silicon MEMS devices,” J. Microelectromech. Syst., vol. 15, no. 5,
vol. 19, no. 1, pp. 192–201, Feb. 2010. pp. 1131–1138, Oct. 2006.
[8] K. Sundaresan, G. K. Ho, S. Pourkamali, and F. Ayazi, “Electronically [31] A. K. Samarao and F. Ayazi, “Post-fabrication electrical trimming of
temperature compensated silicon bulk acoustic resonator reference os- silicon bulk acoustic resonators using joule heating,” in Proc. IEEE Int.
cillators,” IEEE J. Solid-State Circuits, vol. 42, no. 6, pp. 1425–1434, Conf. MEMS, 2009, pp. 892–895.
Jun. 2007.
[9] C. T.-C. Nguyen and R. T. Howe, “Microresonator frequency control and
stabilization using an integrated micro oven,” in Proc. Int. Conf. Solid-
State Sens., Actuators Microsyst. (Transducers), 1993, pp. 1040–1043.
[10] C. M. Jha, M. A. Hopcroft, S. A. Chandorkar, J. C. Salvia, M. Agarwal,
R. N. Candler, R. Melamud, B. Kim, and T. W. Kenny, “Thermal isolation Ashwin K. Samarao (S’07–M’11) was born in
of encapsulated MEMS resonators,” J. Microelectromech. Syst., vol. 17, Chennai, India, on February 18, 1983. He received
no. 1, pp. 175–184, Feb. 2008. the B.E. (Hons.) degree in electrical and computer
[11] J. S. Wang and K. M. Lakin, “Low-temperature coefficient bulk acoustic engineering from the Birla Institute of Technology
wave composite resonators,” Appl. Phys. Lett., vol. 40, no. 4, pp. 308–310, and Science, Pilani, India, in 2004, the M.S. degree
Feb. 1982. in electrical and computer engineering from the Uni-
[12] R. Tabrizian, G. Casinovi, and F. Ayazi, “Temperature-stable high-Q AlN- versity of Cincinnati, Cincinnati, OH, in 2006, and
on-silicon resonators with embedded array of oxide pillars,” in Proc. the Ph.D. degree in electrical and computer engineer-
Solid-State Sens., Actuators Microsyst. Workshop (Hilton Head), 2010, ing from the Integrated Microelectromechanical Sys-
pp. 100–101. tems Laboratory, Georgia Institute of Technology,
[13] R. Melamud, B. Kim, S. A. Chandorkar, M. A. Hopcroft, M. Agarwal, Atlanta, in May 2011.
C. M. Jha, and T. W. Kenny, “Temperature-compensated high-stability Subsequently he joined Integrated Device Technology, Inc., San Jose, CA, as
silicon resonators,” Appl. Phys. Lett., vol. 90, no. 24, pp. 244 107-1– an intern, where he worked on implementing the ideas he developed as a gradu-
244 107-3, Jun. 2007. ate student to engineer a MEMS-based timing solution. He is currently a MEMS
[14] R. Abdolvand, H. M. Lavasani, G. Ho, and F. Ayazi, “Thin-film Research Engineer with Robert Bosch Research and Technology Center, Palo
piezoelectric-on-silicon resonators for high-frequency reference oscillator Alto, CA. His research interests include micro- and nanoelectromechanical
applications,” IEEE Trans. Ultrason., Ferroelectr., Freq. Control, vol. 55, resonators, RF-MEMS and inertial sensors, CMOS-MEMS integration, and
no. 12, pp. 2596–2606, Dec. 2008. micro- and nanofabrication process development.
[15] G. Bahl, R. Melamud, B. Kim, S. A. Chandorkar, J. C. Salvia, Dr. Samarao was the recipient of the Best Student Paper Award at the IEEE
M. A. Hopcroft, D. Elata, R. G. Hennessy, R. N. Candler, R. T. Howe, and International Frequency Control Symposium in 2010.
T. W. Kenny, “Model and observations of dielectric charge in thermally
oxidized silicon resonators,” J. Microelectromech. Syst., vol. 19, no. 1,
pp. 162–174, Feb. 2010.
[16] S. Pourkamali, G. K. Ho, and F. Ayazi, “Low-Impedance VHF and UHF
capacitive silicon bulk acoustic wave resonators—Part I: Concept and
fabrication,” IEEE Trans. Electron Devices, vol. 54, no. 8, pp. 2017–2023, Farrokh Ayazi (S’96–M’00–SM’05) received the
Aug. 2007. B.S. degree in electrical engineering from the Uni-
[17] P. Csavinszky and N. G. Einspruch, “Effect of doping on elastic constants versity of Tehran, Tehran, Iran, in 1994 and the M.S.
of silicon,” Phys. Rev. Lett., vol. 132, no. 6, pp. 2434–2440, Dec. 1963. and Ph.D. degrees in electrical engineering from the
[18] W. P. Mason, “Ultrasonic attenuation and velocity changes in doped University of Michigan, Ann Arbor, in 1997 and
n-type germanium and p-type silicon and their use in determining an 2000, respectively.
intrinsic electron and hole scattering time,” Phys. Rev. Lett., vol. 10, no. 5, He joined the faculty of the Georgia Institute of
pp. 151–154, Mar. 1963. Technology, Atlanta, in December 1999, where he
[19] A. K. Samarao and F. Ayazi, “Temperature compensation of silicon mi- is currently a Professor in the School of Electrical
cromechanical resonators via degenerate doping,” in IEDM Tech. Dig., and Computer Engineering. His research interests are
2009, pp. 789–792. in the areas of integrated micro- and nanoelectro-
[20] P. Csavinszky, “Effect of holes on the elastic constant C’ of degenerate mechanical resonators, interface IC design for MEMS and sensors, inertial
p-type Si,” J. Appl. Phys., vol. 36, no. 12, pp. 3723–3727, Dec. 1965. sensors, RF MEMS, and microfabrication techniques.
[21] W. R. Thurber, R. L. Mattis, Y. M. Liu, and J. J. Filliben, “Resistivity- Prof. Ayazi is a 2004 recipient of the NSF CAREER Award, the 2004
dopant density relationship for boron doped silicon,” J. Electrochem. Soc., Richard M. Bass Outstanding Teacher Award (determined by the vote of the
vol. 127, no. 10, pp. 2291–2294, Oct. 1980. ECE senior class), and the Georgia Tech College of Engineering Cutting
[22] C Iliescu, M. Carp, J. Miao, F. E. H. Tay, and D. P. Poenar, “Analysis Edge Research Award for 2001–2002. He is an editor for the IEEE/ASME
of highly doping with boron from spin-on diffusing source,” Surf. Coat. J OURNAL OF M ICROELECTROMECHANICAL S YSTEMS. He served on the
Technol., vol. 198, no. 1–3, pp. 309–313, Aug. 2005. technical program committee of the IEEE International Solid State Circuits
[23] N. Ono, K. Kitamura, K. Nakajima, and Y. Shimanuki, “Measurement Conference for six years (2004–2009). He and his students won the best
of Young’s modulus of silicon single crystal at high temperature and its paper awards at Transducers 2011, the IEEE International Frequency Control
dependency on boron concentration using flexural vibration method,” Jpn. Symposium in 2010, and IEEE Sensors conference in 2007. He is the Co-
J. Appl. Phys., vol. 39, no. 2A, pp. 368–371, Feb. 2000. Founder and Chief Technology Officer of Qualtré Inc., a spin-out from his
[24] J. S. Wang, A. R. Landin, and K. M. Lakin, “Low temperature coefficient research laboratory that commercializes multiaxis bulk-acoustic-wave silicon
shear wave thin films for composite resonators and filters,” in Proc. Ul- gyroscopes and multidegrees-of-freedom inertial sensors for consumer elec-
trasonics Symp., 1983, vol. 1, pp. 491–494. tronics and personal navigation systems.

You might also like