The crystallization behavior of the supercooled bulk metallic glass-forming Zr 41 Ti 14 Cu 12 Ni 10 Be 23 liquid was studied with different heating and cooling rates. A rate of about 1 K/s is sufficient to suppress crystallization of the melt upon cooling from the equilibrium liquid. Upon heating, in contrast, a rate of about 200 K/s is necessary to avoid crystallization. The difference between the critical heating and cooling rate is discussed with respect to diffusion-limited growth taking classical nucleation into account. The calculated asymmetry of the critical heating and cooling rate can be explained by the fact that nuclei formed during cooling and heating are exposed to different growth rates. ͓S0163-1829͑99͒07441-X͔The ability to form a glass by cooling from the equilibrium liquid is equivalent to suppressing crystallization within the supercooled ͑undercooled͒ liquid. One of the central quantities in theoretical and experimental studies of glass formation is the critical cooling rate R c to bypass crystallization upon cooling from the stable melt. 1 Critical cooling rates of monoatomic metallic systems are typically of the order of 10 12 K/s. In contrast, R c of recently discovered multicomponent bulk metallic glass-forming alloys 2,3 is of the order of a few K/s. This excellent glass-forming ability enables investigations of crystallization, 4,5 viscosity, 6 and diffusion 7,8 as well as relaxation 9 in the supercooled liquid region. The critical cooling rate of Zr 41 Ti 14 Cu 12 Ni 10 Be 23 ͑Vit 1͒, studied in this work, is about 1 K/s. 10 In contrast, crystallization of amorphous Vit 1, previously investigated by differential scanning calorimetry ͑DSC͒ upon heating 11 could not be avoided up to the maximum heating rate of the DSC of 5 K/s and the critical heating rate has not been determined yet. The critical heating rate R h , the counterpart of the critical cooling rate upon heating, is the lowest rate an amorphous sample can be heated through the entire supercooled liquid region without crystallization.In this paper, the onset temperature of crystallization is investigated during cooling from the stable melt and heating amorphous Vit 1 as a function of cooling and heating rate, respectively. For this purpose we designed an experimental setup that permits maximum heating rates of 350 K/s and maximum cooling rates of 40 K/s. We will show that an asymmetry of R c and R h results from the fact that nuclei formed during cooling and heating are exposed to different growth rates, which is likely to be a general feature for metallic systems.The investigations were performed in high-purity graphite crucibles since heterogeneous surface nucleation at the container walls does not effect the crystallization of the bulk Vit 1 sample. 12 The samples were mounted into the graphite crucibles and inductively heated in vacuum of 10 Ϫ6 mbar or in a titanium-gettered argon atmosphere. The temperature was measured using a thermocouple ͑type K͒ with an accuracy better than Ϯ2 K. Details of the experimental setup can b...