Studies of nucleosynthesis in neutrino-driven winds from nascent neutron stars show that the elements from Sr through Ag with mass numbers A \sim 88 –110 are produced by charged-particle reactions ( CPR ) during the \alpha -process in the winds .
Accordingly , we have attributed all these elements in stars of low metallicities ( { [ Fe / H ] } \lesssim - 1.5 ) to low-mass and normal supernovae ( SNe ) from progenitors of \sim 8 – 11 M _ { \odot } and \sim 12 – 25 M _ { \odot } , respectively , which leave behind neutron stars .
Using this rule and attributing all Fe production to normal SNe , we previously developed a phenomenological two-component model , which predicts that { [ Sr / Fe ] } \geq - 0.32 for all metal-poor stars .
The high-resolution data now available on Sr abundances in Galactic halo stars show that there is a great shortfall of Sr relative to Fe in many stars with { [ Fe / H ] } \lesssim - 3 .
This is in direct conflict with the above prediction .
The same conflict also exists for two other CPR elements Y and Zr .
The very low abundances of Sr , Y , and Zr observed in stars with { [ Fe / H ] } \lesssim - 3 thus require a stellar source that can not be low-mass or normal SNe .
We show that this observation requires a stellar source leaving behind black holes and that hypernovae ( HNe ) from progenitors of \sim 25 – 50 M _ { \odot } are the most plausible candidates .
Pair-instability SNe from very massive stars of \sim 140 – 260 M _ { \odot } that leave behind no remnants are not suitable as they are extremely deficient in producing the elements of odd atomic numbers such as Na , Al , K , Sc , V , Mn , and Co relative to the neighboring elements of even atomic numbers , but this extreme odd-even effect is not observed in the elemental abundance patterns of metal-poor stars .

If we expand our previous phenomenological two-component model to include three components ( low-mass and normal SNe and HNe ) and use for example , the observed abundances of Ba , Sr , and Fe to separate the contributions from these components , we find that essentially all of the data are very well described by the new model .
This model provides strong constraints on the evolution of [ Sr/Fe ] with [ Ba/Fe ] in terms of the allowed domain for these abundance ratios .
This model also gives an equally good description of the data when any CPR element besides Sr ( e.g. , Y or Zr ) or any heavy r -process element besides Ba ( e.g. , La ) is used .
As the stars deficient in Sr , Y , and Zr are dominated by contributions from HNe , they define the self-consistent yield pattern of that hypothecated source .
This inferred HN yield pattern for the low- A elements from Na through Zn ( A \sim 23 –70 ) including Fe is almost indistinguishable from what we had previously attributed to normal SNe .
As HNe are plausible candidates for the first generation of stars and are also known to be ongoing in the present epoch , it is necessary to re-evaluate the extent to which normal SNe are substantial contributors to the Fe inventory of the Galaxy .
We conclude that HNe are important contributors to the abundances of the low- A elements over the history of the universe .
We estimate that they contributed \sim 24 \% of the bulk solar Fe inventory while normal SNe contributed only \sim 9 \% ( not the usually assumed \sim 33 \% ) .
This implies a greatly reduced role of normal SNe in the chemical evolution of the low- A elements .