362
Appl. Phys. Lett., Vol. 82, No. 3, 20 January 2003
Oh et al.
FIG. 1. ͑a͒ O–Fe stretching mode of FeO on Fe͑001͒; ͑b͒ vibrational loss
spectrum for 1 ML MgO film deposited on Fe͑001͒; ͑c͒ vibrational loss
spectrum for 1 ML MgO on 1 ML FeO-precovered Fe͑001͒; ͑d͒ spectrum
after annealing the sample in ͑c͒ at 770 K for 20 min. All the spectra were
taken in specular geometry with the incident angle of the electron beam was
60° from the surface normal. The primary electron energy was 11 eV.
FIG. 2. ͑a͒ O–Fe stretching mode of FeO on Fe͑001͒; ͑b͒ vibrational loss
spectrum for 3 ML Mg deposited on FeO precovered Fe͑001͒; ͑c͒ 10 h after
depositing 3 ML Mg; ͑d͒ vibrational loss spectrum after annealing the
sample in ͑c͒ at 770 K for 20 min. All the spectra were taken in specular
geometry with the incident angle of the electron beam was 60° from the
surface normal. The primary electron energy was 11 eV.
vibration loss peak. It shows that the ambient oxygen mol-
ecules are so stable that no detectable amount of oxygen
adsorption to Fe͑001͒ occurs at room temperature. When we
repeat the same experiment with the substrate at 770 K, we
can find the oxygen on the surface by both AES and vibra-
tion spectroscopy. Figure 1͑a͒ shows the vibration loss peak
associated with the O–Fe stretching around 414 cmϪ1 as
previously found near the present loss energy by Lu et al.9
fers a direct evidence for the FeO formation on Fe͑001͒ dur-
ing MgO growth through reactive oxidation.
Since the FeO does not form on Fe͑001͒ at room tem-
perature just by the ambient oxygen molecules of 2
ϫ10Ϫ8 Torr, the present observation of FeO layer formed at
the interface during reactive oxidation shows that Mg should
play a catalytic role to promote the oxidation of Fe substrate.
Since Mg͑O͒ is very electropositive ͑negative͒, thermally
evaporated Mg atoms bond to oxygen molecules by donating
Since the backfilled oxygen molecules of
1
ϫ10Ϫ8 Torr, do not form FeO on Fe͑001͒ at room tempera-
ture, we grow MgO film by the reactive oxidation method,
keeping the same oxygen partial pressure with the substrate
at room temperature. Figure 1͑b͒ shows the vibration loss
spectra for nominal 1 ML thick MgO film grown on Fe͑001͒
at room temperature. Two peaks are found around 504 and
623 cmϪ1. The peak position of the two peaks are deter-
mined by fitting the loss spectrum with two Gaussian peaks
after subtracting linear background. The error limit of the
loss energy is Ϯ7 cmϪ1 as estimated from both statistical
fluctuation and uncertainties in curve fitting. Relative inten-
sity of the latter peak to the former grows as the thickness of
the MgO film increases. Hence, the latter peak is expected to
originate from the MgO film. Furthermore, the loss energy of
the latter peak is similar to that for the surface optical pho-
their electrons to the antibonding molecular orbital, 5 of
*
the oxygen molecules, and induce the dissociation of the
oxygen molecules releasing MgO molecules and oxygen at-
oms. Then, the oxygen atoms react readily not only with
other Mg atoms, but also with the Fe substrate to form FeO
at the interface.
Since the inhomogeneous interface is not desirable as a
model system and also degrades the TMR value of MTJ
severely,5 measures to form sharp interface should be con-
ceived. There appears to be a probable solution to remove
the FeO layer at the interface, when we look into the ther-
modynamic behavior of the film. Comparing Fig. 1͑c͒ with
Fig. 1͑d͒, we find that after annealing 1 ML MgO/1 ML
FeO/Fe͑001͒ at 770 K for 20 min, the loss peak associated
with the MgO is strongly enhanced in sacrifice of the FeO
peak. It can be viewed as the transfer of oxygen from FeO to
unoxidized Mg atoms in the MgO film to form MgO mol-
ecules. ͑It implies that the oxygen partial pressure, 2
ϫ10Ϫ8 Torr, is not enough to fully oxidize the deposited
Mg, and there is unoxidized Mg in the MgO film.͒ It hints
that if we supply enough unoxidized Mg atoms to take the
oxygen from FeO, then FeO-free interface would form.
To examine the abovementioned idea, we initially form 1
ML thick FeO layer on Fe͑001͒ ͓Fig. 2͑a͔͒, then, 3 ML thick
Mg film is deposited on the FeO covered surface at room
temperature without backfilling the chamber with oxygen.
Figure 2͑b͒ shows a very broad peak with its centroid
slightly below 500 cmϪ1. Since it is so broad, it seems to be
composed of many different peaks, possibly originating from
non or Fuchs–Kliewer mode of bulk MgO, 651 cmϪ1 10
.
Hence, we assign it to the Fuchs–Kliewer mode of the
strained, ultrathin MgO film on Fe͑001͒. The other loss peak
at 504 cmϪ1 in Fig. 1͑b͒ may originate from the FeO formed
at the interface concomitantly with MgO film as suggested
by Meyerheim et al.4 To examine the idea, we grow 1 ML
FeO on Fe͑001͒ following the recipe as previously men-
tioned and then deposit 1 ML MgO on the top. Correspond-
ing loss spectrum is shown in Fig. 1͑c͒ and looks very simi-
lar to Fig. 1͑b͒; There are two loss peaks and their peak
positions are coincident with those in Fig. 1͑b͒ within the
error limit, Ϯ7 cmϪ1. Then, the peak with its loss energy,
ϳ504 cmϪ1 is considered to be the stretching mode of FeO
under the influence of the MgO overlayer. According to
above peak assignment, the present spectroscopic work of-
pristine FeO layer, Mg covered FeO layer, and MgO over-
128.114.34.22 On: Wed, 03 Dec 2014 14:31:42
Oh
