The effect of load rate, placement angle, and ice type on
ice screw failure
K. Blair, D. Custer
Massachusetts Institute of Technology, Center for Sports Innovation,
Cambridge, MA, USA
S. Alziati, W. Bennett
Cambridge University, Cambridge, UK
ABSTRACT: Accident reports and previous research show that ice screws often fail to
remain anchored in ice when subjected to the typical loads produced by climbing falls.
This research characterized the failure of ice screws placed in ice and pulled to failure in a
tensile tester. The test matrix is comprised of two ice types (water ice and aerated ice),
two loading rates (125 N/s and 12,500 N/s), and three placement angles (–30º, 0º, 30º).
High-speed video provided a qualitative record of the failure. Four results were worthy of
note. The failure loads were significantly below those reported in the literature. Load rate
significantly affects failure load. Angle, load rate, and ice type show a complex
relationship with failure load. Two failure modes were observed, a sudden “popping out”
and a “slicing out.” These results suggest that climbers should minimize loading rate and
place ice screws where the ice is in compression.
Climber lore suggests that ice is a fickle medium for anchors. This reputation is primarily
due to the nature of ice, whose properties change dramatically with both temperature and
morphology. At low temperatures (below –20ºC) ice exhibits brittle behavior, and above –
5ºC its behavior is plastic (Schulson, 1999). This behavior is further complicated by the
many different factors that contribute to ice formation; temperature history, water flow,
sun exposure, snow entrainment, and glacial processes can all combine to produce the
myriad of formations that climbers climb. Although the variable nature of the medium is
one of the aspects that draws ice-climbers to the sport, this variability presents a challenge
to the design, testing, and use of ice screws.
The seminal ice screw studies, conducted by the German and Austrian Alpine Clubs
(DAV/ÖAV) in the early 1980s, show a wide variability in test results and an alarming
number of ice screw models (22 of the 33 tested) that could not sustain an impact force of
10 kN (Schubert & Mägdefrau, 1985; Damisch, 1985). These tests are conducted in the
field on glacier ice and in the lab on ice made by wetting and refreezing crushed ice to
mimic glacier ice. Several comparable loading methods are used, including dynamic drop
tests with a falling weight, “static” pulls by 6 climbers loading a dynamic rope as swiftly
and as heavily as possible, and pulls from a mechanical tester at a pull rate of 20 to 30
mm/min. Loading rates for all tests are on the order of 100 N/sec. An “optimal” placement
angle is determined based on balancing the pull from both the up and down direction –
important for glacier work but less relevant to modern waterfall climbing. These findings
have prompted improvements in the manufacture of ice screws and led to the Union
Internationale des Associations d´Alpinisme (UIAA) Standard 151, which has been
recently subsumed by the European Standard, EN 568. The standard requires ice screws to
sustain a load of 10 kN at a 100 mm/sec tensioning rate in a refrozen, crushed ice slurry at
–18°C with a 0° placement angle. The DAV/ÖAV studies and the resultant EN standard
have improved ice screw safety by defining a standard testing methodology and setting a
10 kN threshold for minimum failure load.
Despite the intervening 20 years of improvements, recent climbing literature suggests
that ice screws continue to make fickle anchors; they fail in both the lab and the field, and
the failure sometimes comes as a surprise or occurs at a surprisingly low force. Ice screw
“pulling out” makes a regular appearance as a contributing cause of climbing accidents.
Climbers who experience ice screw failure often express consternation at the failure of
ostensibly solid placements. Research shows a wide variation in the failure force
(Harmston, 1997; Luebben, 1997). Typical standard deviations are on the order of 25% of
the average (~20 kN, 5 kN std) implying that 1 in 10 ice screws fail at less than half the
average and about 1 time out of 100 an ice screw will fail below 5 kN, well within the
range of nominal forces produced during a climbing fall (Pavier, 1998) These recent
results continue to emphasize the degree of variability in ice screw performance.
The variability exhibited, even in controlled lab situations, is further exacerbated by
conditions in the field and the choices faced by climbers who must decide how and how
often to place anchors. The Harmsten and Luebben studies address the relationship
between failure load and two variables that climbers can control: ice screw length and
placement angle. The results for both variables run counter to conventional climbing
practice; ice screw length has little effect on failure load and a positive placement angle
correlates with higher failure load. Further elements that a climber might take into
consideration are: the size of the ice feature in which the screw is placed, the structure and
morphology of the ice, the internal stress state of the ice, the temperature, and the
magnitude and duration of a potential fall. To our knowledge, conventional wisdom for all
these situations has not been verified experimentally.
Our work repeated previous studies of placement angle and investigated the effect of
two additional variables: ice type and strain rate. Placement angle was addressed because
of the discrepancy between the previous studies and conventional climber practice. Ice
type was addressed because of the discrepancy between the many types of ice found in the
field and the few types that have been used in previous tests. Strain rate was deemed
important because of the expected relationship between temperature and strain rate for
materials like ice in which fracture is mediated by defect mobility; thus correlating the
effects of fast loading rate are expected to correlate with low temperature.
Two further departures from previous work were introduced to perform the
experiments in a safe and timely fashion. We have used 13 cm ice screws. The use of a
short screw allows us to focus on the failure of the ice; the ice screw is not damaged in the
test. Additionally, the 13 cm screw permitted a test on both ends of an ice cylinder whose
dimensions were limited by the size of the tensile tester and the strength of the
experimenters’ lower spines. Concern for our backs also drove our decision to make our
ice cylinders in PVC tubes rather than in the steel containers used in previous works.
This research characterized the failure of forty-three, 13 cm ice screws placed in ice and
pulled to failure in a tensile tester. The test matrix is comprised of two ice types (water ice
and aerated ice), two loading rates (125 N/s to 12500 N/s), and three placement angles (–
30º, 0º, 30º). A positive angle indicates the ice screw is being tensioned along its length; a
negative angle indicates compression.
In order to test two types of ice, we needed to be able to make two types of ice in a timely
and repeatable manner. We chose a “water” ice, frozen tap water, and an “aerated” ice,
made by freezing a 3:1 mixture of tap water and carbonated water (seltzer water). The
water ice was chosen because it closely mimics both field conditions and the ice in
previous studies. The aerated ice proved simple to make and closely mimicked aerated ice
found in the field; its appearance and “feel” when placing an ice screw were similar to ice
that has either been formed during a snow event or been exposed to solar radiation for
several days at temperatures low enough to preclude melting. The two ice types were
characterized by measuring the density and compressive strength of cylindrical “puck”
samples 16 cm in diameter. Density was determined by measuring volume and weight,
and compressive strength was determined by crushing the samples between steel plates in
an MTS testing machine at a fixed displacement rate of 0.508 mm/s. For ice screw testing,
both ice types were produced in PVC pipe “buckets,” cylindrical containers 30 cm in
diameter (internal) and 40 cm in height; the PVC wall thickness was 15 mm. Filled PVC
cylinders were placed in a freezer at –18ºC; a thin, flexible tube allowed water from the
center to vent as the ice formed. Fig. 1 shows both a “puck” sample of aerated ice that has
been dyed with food coloring to identify the air pockets and a PVC cylinder of water ice
with vent tube.
Fig. 1 A“puck” sample of aerated ice that has been dyed with food coloring identifying
the air pockets (a), and a PVC cylinder of water ice with vent tube (b).
ICE SCREW TESTING
We used a 13 cm long version of a popular brand of ice screw. Ice screws were kept in the
freezer with the ice samples (-18ºC), screwed into ice cylinders within minutes of testing,
and pulled to failure in an MTS tensile tester. Data were recorded at 500 Hz at an
accuracy of ±5 N. Our two pull rates were 25 mm/s (fast) and 0.25 mm/sec (slow), which
translated to loading rates of 125 N/s to 12500 N/s respectively, rates that are
representative of the fastest and slowest loading rates expected in field. Eight tensile tests
were recorded on high-speed video to provide a qualitative, visual record of the failure.
One of the high-speed video test cases was run with 22 cm ice screw to determine whether
these longer ice screws fail in a fashion similar to the failure mode of the 13 cm version.
Our water ice pucks had a mean density of 0.913 g/cm3 with a standard deviation of
0.0008 g/cm3; aerated ice pucks had a mean density of 0.804 g/cm3 with a standard
deviation of 0.03 g/cm3. Compression tests showed similar mean compressive strengths
for both water ice, 0.487 MN/m
2, and aerated ice, 0.398 MN/m
2; large standard deviations
(± 30%) and insignificant F-test results from this measurement suggest that these
compressive strengths did not differ statistically. The similarity between the compressive
behaviors of these two ice types is shown in Fig. 2. A difference was apparent when
observing the compression; after the tests were performed on water ice, the specimens
froze to both the metal plates in the MTS machine and the fingers of the experimenters,
whereas aerated ice was wet to the touch.
Fig. 2 Typical compressive failure results
for water ice (WI) and aerated (AI) ice. Fig. 3 Typical load/displacement curves.
The peak values taken from graphs like this
one are used to produce Fig. 4.
ICE SCREW LOADING
Typical load vs. displacement results are shown in Fig. 3. Once the initial slack has been
taken out of the system, loading proceeded at constant rate. The failure load was taken as
the first peak and the resulting data points are displayed in Fig. 4. In Fig. 4(a), lines are
added to illustrate trends. The same data is shown again in Fig. 4(b) along with the data
from Harmsten, Luebben and DAV/ÖAV for comparison.
-40 -30 -20 -10 010 20 30 40
Placement Angle, Degrees
Failure Load, kN
-40 -30 -20 -10 010 20 30 40
Placement Angle, Degrees
Failure Load, kN
Fig. 4 (a) Failure load vs. placement angle data points for the 4 pairings of water/aerated
and fast/slow load rate. WI is water ice, AI is aerated ice. F and S are fast and slow load
rate. (b) Data from the Harmsten, Luebben and DAV/ÖAV tests are shown for
comparison. Note that data points at -30º, 0º and +30º are expanded ±2º for clarity.
High-speed video footage provided evidence of a qualitative difference between the
failures modes of the fast and slow load rates. Single frame images that portray fast and
slow load rate behavior are shown in Fig. 5 along the same time line. Fast loading rates
caused the ice screw to swiftly “slice” its way out of the ice. The failure mode at slow
rates was characterized by the ice screw suddenly “popping” out of the ice. These failure
modes are reminiscent of failure modes in wood fasteners: axial pullout of screws and
shearing failure in wood connectors.
0 1 2 3-1 0 1 2 3-1
Fig. 5 Failure images from high-speed video. The timeline shown is in seconds. Failure
initiates at t = 0. Fast load rate is shown on the top, slow load rate on the bottom.
We have further characterized the performance of ice screws by investigating how
placement angle, loading rate, and ice type affect failure load and failure mode. Four
trends are worthy of analysis: our low failure loads, the effect of loading rate, the complex
relationship between angle, ice type, loading rate, and failure load, and the two failure
modes observed in the high-speed video recordings.
The most significant result is our low failure loads. One possible reason for these low
loads is our use of PVC rather than steel housing. Our first concern was that the ice screws
were propagating cracks to the outer edge of the cylinder, but our spent cylinders
exhibited no such fractures (with the exception of surface spaulding). Nor were such
fractures visible in the high-speed video recordings. Because cracks do not extend to the
edges, we hypothesize that the size of our ice cylinders is sufficient; this suspicion is
supported by the 5 diameters rule of thumb for the cone of influence for fastening screws.
Our explanation for the low failure loads is that two aspects of our procedure affected the
stress condition of the ice: we used a venting tube to allow water to escape as the freezing
progressed and the PVC allowed substantial expansion. In contrast, the ice constrained in
the steel containers used in previous works cannot expand or escape, resulting in an
internal stress condition absent in our tests. If this hypothesis is valid, then ice screw
strength depends on the internal stress condition of the ice. Further research is needed to
confirm the dependence of failure load on the compressive state of the ice. In the
meantime, climbers should consider preferentially placing ice screws in places where
compression is likely, such as at the bases of columns, rather than at the top of curtains.
Our second most significant result is that a faster loading rate reduces failure load. A
two order of magnitude difference in load rate results in a three-fold decrease in failure
load. Loading rate is a factor that climbers can control through the use of skinny ropes,
soft belays, and closely spaced ice screws. We suspect these precautions become more
relevant as temperature decreases, effectively raising the strain rate. Further work should
be done to determine the relationship between temperature and failure load.
Our third result is the complex relationship between placement angle, ice type, loading
rate, and failure load. While all of our tests of negative placement angle were in agreement
with the previous study results showing reduced failure load, our results for positive
placement angles showed a complicated dependency on loading rate and ice type. Our
data suggests a positive placement angle in aerated ice at a slow loading rate may result in
a very low failure load. Climbers might consider 0º placement angles.
Finally, high-speed video shows two failure modes, which are previously unreported.
At the slow load rate, the ice screw simply “pops out” of the ice, displacing only a little
ice in the immediate vicinity of the ice screw. At the fast load rate, the ice screw “slices” a
slot in the ice as it pulls out. These failure modes are very similar to those reported for
wood fasteners (American Forest). Further refinements of ice screw design might increase
the failure load under these distinct failure modes.
Overall, our research suggests that ice screw behavior is complicated, that it is
sensitive to load rate, placement angle, temperature, internal ice stress state, and ice type,
and that it is subject to two failure modes that correlate closely with load rate. Further
study should address the relationship between failure load and both internal stress and
temperature of the ice.
American Forest & Paper Association (1999) General Dowel Equations for Calculating
Lateral Connection Values, Technical Report 12.
Damisch C. (1985) “Ergänzende Untersuchung con Eisschrauben und Eishaken”,
Sicherheit im Firn und Eis, 243-249, edited Pit Schubert & Helmut Mägdefrau,
German Alpine Club, Munich.
Harmston C. (1997) Myths, Cautions & Techniques of Ice Screw Placement,
Luebben C. (1997) The Cold Truth. Climbing Magazine, Nov 1-Dec 13, 1997, pp. 106-
Pavier M. (1998) Experimental and Theoretical Simulations of Climbing Falls, Sports
Engineering, 1, pp. 79-91.
Schubert P. & Mägdefrau H. (1985) “Untersuchung von Eisschrauben und Eishaken”,
Sicherheit im Firn und Eis, 201-242, edited Pit Schubert & Helmut Mägdefrau,
German Alpine Club, Munich.
Schulson E. M. (1999) The Structure and Mechanical Behavior of Ice. JOM, 51(2) pp. 21-