1999 Macmillan Magazines Ltd

can momentarily produce a single-crack state. As explained in ref.

13, the high peaks in the velocity measurements correspond to these
single-crack states, because, when several cracks are propagating
simultaneously, the available energy is distributed between them,
and the front velocity drops. Thus beyond v
c
there are moments
when we measure the velocity of a single crack moving in a known
energy density.
An important prediction of equation (2) is that a crack, having no
`inertia', will immediately jump to its predicted velocity. Another
prediction of linear elastic theory
12
, veried in PMMA
18
, is that
upon a sudden change in loading, the stress eld will assume its
asymptotic value within the time needed for a shear wave to pass.
Thus, ,1 ms after the `death' of a side branch of length 1 mm, the
stress eld throughout the singular zone will be that of a single
crack. As above 450 ms
-1
, the system becomes effectively two-
dimensional (ref. 2), the highest peaks of the instantaneous velocity
should then be described by equation (2); alternatively, we can use
equation (1) to derive the value of for a single crack moving at the
peak velocities. We check this premise by comparing these derived
values with the measurements of in a strip
2,14
. These measure-
ments showed that above v
c
, when the total fracture surface formed
by both the main crack and its branches was taken into account,
(per unit fracture surface) has the constant value of 0.9(v
c
).
Comparison between the measured and derived values of was
performed using the peak velocities obtained in a number of
experiments that were conducted under widely diverse conditions.
The results (Fig. 3a) indicate that the theory works remarkably well.
As in ref. 14, the derived values of (v) were constant and equal to
0.9(v
c
). The 10% drop in relative to (v
c
) may result from the
extremely high acceleration rate preceding the peak velocities. This
effect also appears at fracture initiation in PMMA, when at comparable
acceleration rates, an overshoot of the initial crack velocity,
consistent with an effective 10% reduction in , is observed.
In Fig. 3b, c we directly compare the instantaneous velocity
measurements of a number of different cracks with the correspond-
ing velocity curves predicted by equation (2). In all of the data sets
the theoretical curve was calculated (with no adjustable parameters)
using a single value of v <0:9v
c
(= 3,000 J m
-2
in PMMA).
Above v
c
, all of the highest-velocity peaks are accurately described
by the theoretical curve, despite the wide variations in experimental
conditions. At the highest energy uxes, peak velocities, correctly
described by equation (2), indeed exist in excess of 0.9v
R
. Thus to see
the approach to v
R
, one need only look at the instantaneous velocity
of single-crack states. These results are consistent with both
calculations
8
and experiments
19
in which the micro-branching
instability was articially suppressed. M
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Methods
The experiments were conducted in thin, quasi-two-dimensional cast PMMA
and soda lime glass sheets of size 380 3440 mm in the x (propagation) and y
(loading) directions, and thickness 2 and 3 mm, respectively. All samples were
loaded by uniform displacement of the vertical boundaries. Fracture was
initiated at the tip of a small notch inserted midway between the vertical
boundaries at the sample's edge. The initial length and tip curvature of this
notch determined the amount of elastic energy stored in the sample before the
onset of fracture. The crack velocity was measured by the technique described
in ref. 2 with the addition of analogue differentiation of the signal before its
digitization. A velocity resolution of better than 10 ms
-1
in PMMA and of the
order of 50 ms
-1
in glass was attained. All the data presented were measured
before the arrival of reected stress waves from the far boundaries of the plates.
Reected waves from the sample edge behind the initial crack, as in ref. 20, had
no noticeable effect on the crack. Thus the loading conditions were, effectively,
those of constant stress loading. As this can be mapped to constant loading of
the crack faces, the experiments are compatible with the assumptions leading to
equation (1). G(l) was computed numerically for the precise geometry of
the plates we used. The values of v
R
, obtained by a direct measurement, are
930 ms
-1
(PMMA) and 3,300 ms
-1
(glass).
Received 9 July; accepted 5 October 1998.
1. Bergkvist, H. Some experiments on crack motion and arrest in polymethylmethacrylate. Eng. Fracture
Mech. 6, 621626 (1974).
2. Sharon, E. & Fineberg, J. Microbranching instability and the dynamic fracture of brittle materials.
Phys. Rev. B 54, 71287139 (1996).
3. Ravi-Chandar, K. & Knauss, W. G. An experimental investigation into dynamic fracture. III. On
steady-state crack propagation and crack branching. Int. J. Fracture 26, 141154 (1984).
4. Ravi-Chandar, K. & Knaus, W. G. An experimental investigation into dynamic fracture. II.
Microstructural aspects. Int. J. Fracture 26, 6580 (1984).
5. Fineberg, J., Gross, S. P., Marder, M. & Swinney, H. L. Instability in the propagation of fast cracks.
Phys. Rev. B 45, 51465154 (1992).
6. Marder, M. & Liu, X. M. Instability in lattice fracture. Phys. Rev. Lett. 71, 24172424 (1993).
7. Marder, M. & Gross, S. Origin of crack tip instabilities. J. Mech. Phys. Solids 43, 148 (1995).
8. Xu, X. P. & Needleman, A. Numerical simulations of fast crack growth in brittle solids. J. Mech. Phys.
Solids 42, 13971437 (1994).
9. Johnson, E. Process region changes for rapidly propagating cracks. Int. J. Fracture 55, 4763 (1992).
10. Lobkovsky, A. & Langer, J. S. Critical examination of cohesive-zone models in the theory of dynamic
fracture. J. Mech. Phys. Solids 46, 15211526 (1998).
11. Ching, E. S. C., Langer, J. S. & Nakanishi, H. Dynamic instabilities in fracture. Phys. Rev. Lett. 76,
10871090 (1996).
12. Freund, L. B. Dynamic Fracture Mechanics (Cambridge Univ. Press, New York, 1990).
13. Sharon, E., Gross, S. P. & Fineberg, J. Local crack branching as a mechanismfor instability in dynamic
fracture. Phys. Rev. Lett. 74, 50965099 (1995).
14. Sharon, E., Gross, S. P. &Fineberg, J. Energy dissipation in dynamic fracture. Phys. Rev. Lett. 76, 2117
2120 (1996).
15. Boudet, J. F., Ciliberto, S. &Steinberg, V. Dynamics of crack propagation in brittle materials. J. Phys. II
6, 14931516 (1996).
16. Abraham, F. F., Brodbeck, D., Rafey, R. A. & Rudge, W. E. Instability dynamics of fracturea
computer simulation investigation. Phys. Rev. Lett. 73, 272275 (1994).
17. Zhou, S. J., Lomdahl, P. S., Thomson, R. & Holian, B. L. Dynamic crack processes via molecular
dynamics. Phys. Rev. Lett. 76, 23182321 (1996).
18. Vu, B. Q. &Kinra, V. K. Brittle fracture of plates in tension. Static eld radiated by a suddenly stopping
crack. Eng. Fracture Mech. 15, 107114 (1981).
19. Washabaugh, P. D. & Knauss, W. G. A reconciliation of dynamic crack velocity and rayleigh wave
speed in isotropic brittle solids. Int. J. Fracture 65, 97114 (1994).
20. Morrissey, J. W. & Rice, J. R. Crack front waves. J. Mech. Phys. Solids 46, 467487 (1998).
Acknowledgements. We thank L. B. Freund for supplying us with nite-element calculations of G(l). This
work was supported by the US-Israel Binational Fund.
Correspondence and requests for materials should be addressed to J.F. (e-mail: [email protected]).
letters to nature
NATURE | VOL 397 | 28 JANUARY 1999 | www.nature.com 335
Acontrolled-releasemicrochip
John T. Santini Jr*, Michael J. Cima & Robert Langer*
* Department of Chemical Engineering, Department of Materials Science
and Engineering, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, USA
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Much previous work in methods of achieving complex drug-
release patterns has focused on pulsatile release from polymeric
materials in response to specic stimuli
1
, such as electric
25
or
magnetic
6,7
elds, exposure to ultrasound
7,8
, light
9
or enzymes
10
,
and changes in pH
11
or temperature
1214
. An alternative method
for achieving pulsatile release involves using microfabrication
technology to develop active devices that incorporate micro-
metre-scale pumps, valves and ow channels to deliver liquid
solutions
15,16
. Here we report a solid-state silicon microchip that
can provide controlled release of single or multiple chemical
substances on demand. The release mechanism is based on the
electrochemical dissolution of thin anode membranes covering
microreservoirs lled with chemicals in solid, liquid or gel form.
We have conducted proof-of-principle release studies with a
prototype microchip using gold and saline solution as a model
electrode material and release medium, and we have demon-
strated controlled, pulsatile release of chemical substances with
this device.
Controlled release from our microchip involves no moving parts.
Release froma particular reservoir is initiated by applying an electric
potential between the anode membrane covering that reservoir and
a cathode. Fig. 1a shows a cut-away portion of a prototype
microchip containing reservoirs lled with the chemical to be
released. The devices used in this study were 17 mm by 17 mm by
310 mmand contained 34 reservoirs. Device size could be reduced to
,2 mm, depending on the particular application. As a point of
reference, a device of the size used in these studies (17 mm) has
enough surface area to accommodate over 1,000 reservoirs.
1999 Macmillan Magazines Ltd
letters to nature
336 NATURE | VOL 397 | 28 JANUARY 1999 | www.nature.com
Devices for this study were fabricated by a sequential process
using silicon wafers and microelectronic processing techniques
17,18
including ultraviolet photolithography, chemical vapour deposition
(CVD), electron beam evaporation and reactive ion etching. Each
device contained reservoirs that extended completely through the
wafer. The reservoirs were square pyramidal in shape (that is, they
contained one large and one small square opening, see Fig. 1b), had
a volume of ,25 nl, and were sealed on the small square end
(50 350 mm) by a 0.3-mm-thick, gold membrane anode. Gold
was chosen as a model membrane material because it is easily
deposited and patterned, has a low reactivity with other substances
and resists spontaneous corrosion in many solutions over the entire
pH range. However, the presence of a small amount of chloride ion
creates an electric potential region which favours the formation of
soluble gold chloride complexes
19
. Holding the anode potential in
this corrosion region enables reproducible gold dissolution. Poten-
tials below this region are too low to cause appreciable corrosion,
whereas potentials above this region result in gas evolution and
formation of a passivating gold oxide layer that causes corrosion to
slow or stop. Other metals such as copper or titanium tend to
dissolve spontaneously under these conditions or do not form
soluble materials on application of an electric potential. Although
it is used as a model compound in these studies, gold has also been
shown to be a biocompatible material
20
.
Some portions of the electrodes must be protected from
unwanted corrosion by an adherent, non-porous coating that
isolates the electrode materials from the surrounding electrolyte.
Silicon dioxide (SiO
2
) was used as a model protective coating
because its physical properties can be tailored to a particular
application by selecting appropriate processing conditions. For
example, SiO
2
deposited by CVD at low temperatures (,100 8C)
tends to be porous and non-adherent in solution, whereas high-
temperature (.700 8C) deposition or annealing results in denser
SiO
2
but may also lead to thermal grooving and void formation in
gold lms
21
. We used plasma-enhanced CVD at moderate tempera-
tures (350 8C) to address these problems and produce SiO
2
lms
possessing adequate density and adhesion with negligible gold void
formation. Light microscopy revealed no evidence of corrosion
following the applied potential in those areas of the gold electrodes
covered by SiO
2
, except in those regions immediately adjacent to the
exposed gold membrane.
The gold membrane anodes of each microchip were examined for
defects (such as pinholes) using reected and transmission light
microscopy after completion of the entire fabrication process.
Reservoirs covered by defect-free, gold membrane anodes were
considered viable for use in chemical-release experiments and
were loaded with model chemical substances, sodium uorescein
and
45
Ca
2+
(as CaCl
2
), selected for ease of detection. Inkjet printing,
in combination with a computer-controlled alignment apparatus
22
,
is capable of depositing as little as 0.2 nl of a liquid or gel solution of
known concentration into each reservoir. Microsyringe pumps also
have the ability to deposit nanolitre quantities of solutions into
device reservoirs. We found that these two lling methods work well
and prevent any leaking of solutions or gels, such as our aqueous
solutions composed of 1525% (by volume) liquid polymer (poly-
ethylene glycol, relative molecular mass M
r
200) and the model
release compounds. The water quickly evaporates after injection
leaving only liquid polymer and model compound in the reservoir.
The reservoirs are then covered with squares of a thin adhesive
plastic and sealed with a waterproof epoxy. The gold membranes do
not appear to be fragile even though the liquid material in a lled
reservoir can exert a stress on the gold membrane anode due to
capillary forces. Calculations indicate that these capillary forces are
nearly two orders of magnitude smaller than that required to
rupture a 0.3-mm-thick gold membrane, which can withstand
pressures up to 3 lb in
-2
. In addition, prototype devices have been
stored for over a year without any known effects on operation or
reliability.
The objective of initial release experiments was to determine if
pulsatile release of a single compound could be obtained from a
microchip device. Release was achieved from a reservoir of a
prototype device immersed in phosphate-buffered saline (PBS) by
applying a potential of +1.04 volts with respect to a saturated
a
b
Figure 1 A prototype microchip for controlled release showing the shape of a
single reservoir. a, Fabrication of these microchips began by depositing ,0.12 mm
of lowstress, silicon-rich nitride on both sides of prime grade, (100) silicon wafers
using a vertical tube reactor. The silicon nitride layer on one side of the wafer was
patterned by photolithography and electron cyclotron resonance (ECR)
enhanced reactive ion etching (RIE) to give a squaredevice (17 317 mm) contain-
ing 34,480-mm-square reservoirs. The silicon nitride served as an etch mask for
potassium hydroxide solution at 85 8C, which anisotropically etched square
pyramidal reservoirs (b) into the silicon along the (111) crystal planes until the
silicon nitride lm on the opposite side of the wafer was reached. The newly
fabricated silicon nitride membranes completely covered the 4060 mm square
openings of the reservoir. Gold electrodes (0.3 mmthick, with a 0.01-mmchromium
adhesion layer) were deposited and patterned over the silicon nitride membranes
by electron beam evaporation and lift-off. A 0.6-mm layer of plasma enhanced
chemical vapour deposition silicon dioxide was deposited at 350 8C over the
entire electrode containing surface. The silicon dioxide located over portions of
the anode, cathode and bonding pads were etched with ECR-enhanced RIE to
expose the underlying gold lm; this technique was then used to remove the thin
silicon nitride and chromiummembranes located in the reservoir underneath the
gold anode.
a b
Figure 2 Removal of an anode membrane to initiate release from a reservoir. a, b,
Scanning electron micrographs of a gold membrane anode covering a reservoir
are shown before (a) and after (b) the application of +1.04 V w.r.t. SCE for several
seconds in PBS. (Scale bar, 50 mm.)
1999 Macmillan Magazines Ltd
letters to nature
NATURE | VOL 397 | 28 JANUARY 1999 | www.nature.com 337
calomel reference electrode (+1.04 Vw.r.t. SCE) to the gold anode
covering that reservoir. The use of a reference electrode ensured that
the potential of the gold membrane anode remained in the gold
corrosion region during activation. Figures 2a and b show a gold
membrane anode before and after a potential was applied in PBS.
Release of the uorescent model compound, sodium uorescein,
from a reservoir was detected by uorescence spectroscopy within
12 minutes after the potential was applied (Fig. 3).
Subsequent release experiments were designed to determine if the
independent release of multiple compounds could be obtained from
a single device. Reservoirs containing one of two model compounds
were opened by applying +1.04 Vw.r.t. SCE to the corresponding
anode immersed in saline solution without phosphate buffer. The
pulsatile release of
45
Ca
2+
ions and sodium uorescein over a period
of several hours was achieved, showing that multiple compounds
can be released independently from a single microchip device
(Fig. 4).
The release studies demonstrate that the activation of each
reservoir can be controlled individually, creating a possibility for
achieving many complex release patterns. Varying amounts of
chemical substances in solid, liquid or gel form can be released
into solution in either a pulsatile manner, a continuous manner, or a
combination of both, either sequentially or simultaneously from a
single device. Such a device has additional potential advantages
including small size, quick response times and low power con-
sumption. In addition, all chemical substances to be released are
stored in the reservoirs of the microchip itself, creating a possibility
for the future development of autonomous devices. A microbattery,
multiplexing circuitry and memory could be integrated directly
onto the device, allowing the entire device to be mounted onto the
tip of a small probe, implanted, swallowed, integrated with micro-
uidic components to develop a `laboratory-on-a-chip', or incor-
porated into a standard electronic package, depending on the
particular application. Some elds where this device may become
useful include medical diagnostics, analytical chemistry, chemical
detection, industrial process monitoring and control, combinatorial
chemistry and microbiology. Future work will focus on improving
reservoir lling methods and determining the factors that control
the release of material fromopened reservoirs. Gold and siliconwere
used as model materials for the prototype device in these proof-of-
principle release studies. In the future, proper selection of biocom-
patible device materials may result in the development of an
autonomous, controlled-release implant (`pharmacy-on-a-chip')
or a highly controllable tablet (`smart tablet') for drug delivery
applications. M
Received 25 August; accepted 16 November 1998.
1. Kost, J. & Langer, R. Responsive polymeric delivery systems. Adv. Drug Delivery Rev. 6, 1950 (1991).
2. Kwon, I. C., Bae, Y. H. & Kim, S. W. Electrically erodible polymer gel for controlled release of drugs.
Nature 354, 291293 (1991).
Figure 3 Pulsatile release of a single substance from a microchip device. a, The
total mass of sodium uorescein released into PBS over a period of several days
is shown for each of four reservoirs. This release study was conducted in PBS
stirred with a magnetic stirring bar at room temperature. The device was
submerged in the PBS for .36 h before the rst release to ensure that there was
no leakage fromany of the loaded reservoirs. The tipof the SCEwas placedwithin
13 mm of the anode surface immediately before application of the electric
potential. Release of uorescein was obtained on demand by applyinga potential
of +1.04 V w.r.t. SCE for up to 30 s to each anode. The SCE was removed
immediately after the potential was applied. Samples (3 ml) were taken every few
minutes from the release medium, analysed for dye content in a uorimeter, and
replaced with an equal volume of fresh PBS solution. The entire release medium
(100 ml) was replaced with fresh solution several hours after each release. The
release rate for the same experiment is shown in b. The release rate was
calculated by taking the amount of uorescein released between two time points
and dividing it by the time elapsed, starting after the device was immersed in the
PBS.
50
40
30
20
10
0
R
e
l
e
a
s
e

r
a
t
e
0 10 20 30 40 50 60 70
Time (h)
Figure 4 Pulsatile release of multiple substances from a single microchip device.
The release rate of
45
Ca
2+
ions (open triangles; vertical scale in units of
5 3nCi min
21
) and sodium uorescein (lled circles; in units of ngmin
-1
) into
0.145 MNaCl solution over several hours is shown for each reservoir. The experi-
mental procedures and release-rate calculations were the same as in Fig. 3, with
the additionthat samples (1 ml) weretakenevery several minutes fromthe release
medium, analysed for radionuclide content in a scintillation counter, and replaced
with an equal volume of fresh saline solution.
1999 Macmillan Magazines Ltd
letters to nature
338 NATURE | VOL 397 | 28 JANUARY 1999 | www.nature.com
3. Bae, Y. H., Kwon, I. C. & Kim, S. W. in Polymeric Drugs and Drug Adminsitration (ed. Ottenbrite, R.
M.) 98110 (Am. Chem. Soc., Washington DC, 1994).
4. Miller, L. L. Electrochemically controlled release of drug ions from conducting polymers. Mol. Cryst.
Liq. Cryst. 160, 297301 (1988).
5. Hepel, M. & Fijalek, Z. in Polymeric Drugs and Drug Adminsitraiton (ed. Ottenbrite, R. M.) 7997
(Am. Chem. Soc., Washington DC, 1994).
6. Edelman, E. R., Kost, J., Bobeck, H. & Langer, R. Regulation of drug release frompolymer matrices by
oscillating magnetic elds. J. Biomed. Mater. Res. 19, 6783 (1985).
7. Kost, J. & Langer, R. in Pulsed and Self-Regulated Drug Delivery (ed. Kost, J.) 316 (CRC, Boca Raton,
1990).
8. Kost, J., Leong, K. & Langer, R. Ultrasound-enhanced polymer degradation and release of incorpo-
rated substances. Proc. Natl Acad. Sci. USA 86, 76637666 (1989).
9. Mathiowitz, E. & Cohen, M. D. Polyamide microcapsules for controlled release. V. Photochemical
release. J. Membr. Sci. 40, 6786 (1989).
10. Fischel-Ghodsian, F., Brown, L., Mathiowitz, E., Brandenburg, D. & Langer, R. Enzymatically
controlled drug delivery. Proc. Natl Acad. Sci. USA 85, 24032406 (1988).
11. Siegel, R. A., Falamarzian, M., Firestone, B. A. & Moxley, B. C. pH-controlled release from
hydrophobic/polyelectrolyte copolymer hydrogels. J. Control. Release 8, 179182 (1988).
12. Bae, Y. H., Okano, T., Husu, R. & Kim, S. W. Thermo-sensitive polymers as on-off switches for drug
release. Makromol. Chem. Rapid Commun. 8, 481485 (1987).
13. Hoffman, A. S., Afrassiabi, A. & Dong, L. C. Thermally reversible hydrogels: II. Delivery and selective
removal of substances from aqueous solutions. J. Control. Release 4, 213222 (1986).
14. Okano, T., Bae, Y. H. & Kim, S. W. in Pulsed and Self-Regulated Drug Delivery (ed. Kost, J.) 1745
(CRC, Boca Raton, 1990).
15. Gravesen, P., Branebjerg, J. & Jensen, O. S. Microuidics - a review. J. Micromech. Microeng. 3, 168
182 (1993).
16. Shoji, S. & Esashi, M. Microow devices and systems. J. Micromech. Microeng. 4, 157171 (1994).
17. Wolf, S. & Tauber, R. N. Silicon Processing for the VLSI Era Vol. 1 Process Technology (Lattice Sunset
Beach, California, 1986).
18. Jaeger, R. C. Introduction to Microelectronic Fabrication (Addison-Wesley, Reading, Massachusetts, 1988).
19. Frankenthal, R. P. &Siconol, D. J. The anodic corrosion of gold in concentrated chloride solutions. J.
Electrochem. Soc. 129, 11921196 (1982).
20. Merchant, B. Gold, the noble metal and the paradoxes of its toxicology. Biologicals 26, 4959 (1998).
21. Jiran, E. &Thompson, C. V. Capillary instabilities in thin, continuous lms. Thin Solid Films 208, 23
28 (1992).
22. Wu, B. M. in Microstructural Control During Three Dimensional Printing of Polymeric Medical Devices
Thesis, 2540 MIT (1997).
Acknowledgements. We thank A. Gopferich and M. Llabres for help in the early stages of the project, and
K. Jensen, C. Thompson and R. Latanision for discussions about materials and electrochemical aspects of
this project. All fabrication work was carried out at the Microsystems Technology Laboratory at MIT. This
work was partially supported by the US NSF.
Correspondence and requests for materials should be addressed to R.L. (e-mail: [email protected]).
Evidencefor bromine
monoxideinthefree
troposphereduringthe
Arcticpolar sunrise
C. T. McElroy*, C. A. McLinden & J. C. McConnell
* Environment Canada, Downsview, Ontario, Canada, M3H 5T4
Department of Physics & Astronomy, York University, North York, Ontario,
Canada, M3J 1P3
Department of Earth & Atmospheric Science, York University, North York,
Ontario, Canada, M3J 1P3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
During the Arctic polar springtime, dramatic ozone losses occur
not only in the stratosphere but also in the underlying
troposphere
1
. These tropospheric ozone loss events have been
observed over large areas
2,3
in the planetary boundary layer (PBL)
throughout the Arctic
4,5
. They are associated with enhanced
concentrations of halogen species
1,69
and are probably caused
by catalytic reactions involving bromine monoxide (BrO) and
perhaps also chlorine monoxide (ClO)
1,1012
. The origin of the
BrO, the principle species driving the ozone destruction, is
thought to be the autocatalytic release of bromine from sea salt
accumulated on the Arctic snow pack
10,11,13
, followed by photolytic
and heterogeneous reactions which produce and recycle the
oxide
10,11,14,15
. Satellite observations have shown the horizontal
and temporal extent of large BrO enhancements in the Arctic
troposphere
16,17
, but the vertical distribution of the BrO has
remained uncertain. Here we report BrO observations obtained
from a high-altitude aircraft that suggest the presence of signi-
cant amounts of BrO not only in the PBL but also in the free
troposphere above it. We believe that the BrO is transported from
the PBL into the free troposphere through convection over large
Arctic ice leads (openings in the pack ice). The convective trans-
port also lifts ice crystals and water droplets well above the
PBL
18,19
, thus providing surfaces for heterogeneous reactions
that can recycle BrO from less-reactive forms and thereby main-
tain its ability to affect the chemistry of the free troposphere.
We report measurements of tropospheric BrO obtained with the
composition and photodissociative ux measurement (CPFM)
instrument
20,21
, which ew on board the NASA ER-2 high-altitude
research aircraft during the photochemistry of ozone loss in the arctic
region in summer (POLARIS) campaign. Analysis of the CPFM
measurements using differential optical absorption spectroscopy (see
Methods) yields apparent column densities (ACDs), which represent
the number of BrO molecules per unit area along the effective path
taken by the sunlight belowthe aircraft as it is scattered and reected
by the surface into the CPFM instrument. The ACDs obtained
during a ight on 26 April 1997 were found to be in the range of
715 310
14
cm
22
. Figure 1a shows that the ACDs of BrO along
the ight path, shown in Fig. 2a, vary by a factor of three. Also
shown are the estimated tropospheric ACDs after the stratospheric
component has been removed. Accounting for the expected upper
limit of 10 parts per trillion by volume (p.p.t.v.) in the stratosphere
22
results in only a 15 to 35% reduction in the total ACD. Results from
other ights made on 2 May and 6 May 1997 (the latter are also
shown in Fig. 1a) revealed much lower ACDs of BrO, often below
the detection limit of ,2 310
14
cm
22
. At present, we are unaware of
any signicant differences in the meteorology for these periods.
Vertical column densities (VCDs), the number of BrO molecules
within a vertical column above a unit area of surface, are then
extracted from the ACDs (see Methods). In the following discus-
sion, it is assumed that the tropospheric BrO is located at various
heights ranging fromthe surface up to a maximumheight of 5 km, a
height chosen on the basis of observations of plume heights over
leads in Arctic sea ice
18,19
. Figure 1b shows the VCDof BrOalong the
ight track and Fig. 1c shows the computed BrOmixing ratio for the
different schemes investigated. Assuming that all the BrO resides
within the PBL and that the PBL has a height of 1 km, VCDs of
13 310
14
cm
22
or BrO mixing ratios of 50100 p.p.t.v. are
obtained. This is a factor of 23 larger than the maximum BrO
mixing ratio previously observed within the PBL
6
. The case where
the BrO is assumed to reside above the PBL, in a uniform layer
between 1 and 5 km in the troposphere, resulted in a mixing ratio of
2530 p.p.t.v. Another scheme is a uniformly mixed layer between 0
and 5 kmthat yielded slightly lower mixing ratios. The nal case has
a BrO mixing ratio of 50 p.p.t.v. in the 01 km PBL, typical of what
is required to destroy most of the PBL ozone within a day, if
bromine chemistry alone is considered
10
. If the remainder of the
BrOis assumed to reside in the 15 kmaltitude range, mixing ratios
of ,10 p.p.t.v. are obtained. Figure 1df show that there is no
signicant correlation of ACDs of BrO with latitude, altitude or
solar zenith angle (SZA) respectively, as observed along the ight
track, including the ascent (820 km) and a mid-ight dive down to
15 km. This implies that the bulk of the BrO is below 8 km, the
approximate height of the tropopause for this ight, and is therefore
in the troposphere.
From the results shown above and given current estimates of the
amount of inorganic bromine in the PBL
23
, it would seem that a
substantial fraction of the BrO column resides in the free tropo-
sphere. This then presents two problems: how does it get there and
how is it able to remain there as BrO? An associated problem is how
much ozone destruction should be expected in the free troposphere
as a result of the large amounts of BrO observed.
Bromine atoms, produced by the photolysis of BrO, will rapidly
react with species such as HO
2
, HCHO and C
2
H
2
to produce
relatively inactive HBr. Hence, in the absence of heterogeneous
recycling, the BrO will rapidly disappear. Using the model of Tang