b-bbar final states in Higgs production via weak boson fusion at the LHC

We present a study of the Higgs production at the LHC via Weak Boson Fusion, with the Higgs boson decaying into a b-quark pair. A detailed partonic LO calculation of all the potential backgrounds is performed. We conclude that this channel for Higgs production can be extracted from the backgrounds, and present our estimates of the accuracy in the determination of the Hbb Yukawa coupling.


Introduction
A Higgs boson in the so-called low-mass region (115 < m H (GeV) < 140) decays predominantly in b b final states.Due to the large inclusive QCD backgrounds, detection of this decay is however extremely challenging.In particular, the extraction of the most copious signal, namely inclusive gg → H → b b production, has never been shown to be viable.The only production channels which have so far been proven to be suitable for a determination of the Hb b coupling are the associate production Ht t and HW [1,2].
In this note we document a study of the H → b b decay in the electroweak boson fusion (WBF) production channel and of its backgrounds, and we discuss the potential of this process for the determination of the y Hbb Yukawa coupling.The signal rate is proportional to the product of the y HV V coupling, where V denotes a weak W or Z boson, times the B(H → b b) branching ratio.The contamination to the signal coming from QCD production of Higgs plus two jets (mediated by a loop of virtual top quarks) are not included in this analysis.Following the study of ref. [3], these will be suppressed by the particular set of kinematical cuts chosen in our analysis (see Section 2).
The results obtained are based on a leading order partonic calculation of the matrix elements (ME) describing signal and background processes.The latter include the following channels: QCD b bjj production, Z(→ b b)jj, W/Z(→ jj)b b, t t → b b + jets, QCD four jets production (where two light jets are misidentified as generated by b quarks), and contributions from multiple overlapping events.
We identify a set of kinematical cuts leading to signal significances in the range of 2 − 5σ, depending on the Higgs mass.In the lowest mass region, this provides a determination of the B(H → b b) branching ratio with a precision of the order of 20%.The H → b b decay in the WBF channel could be used together with other processes already examined in literature for a model independent determination of the ratio of Yukawa couplings y Hbb /y Hτ τ [4].We therefore conclude that the H → b b channel produced in association with two jets should be considered as an additional channel to be exploited for interesting measurements of the Higgs couplings to fermions.This letter is organized as follows.In Section 2 we describe the kinematical constraints introduced to perform the event selection.Section 3 is devoted to the discussion of signal and backgrounds, while the signal significance and the accuracy of the branching ratio H → b b and Yukawa coupling determination are presented in Section 4. In the Conclusions we summarise and discuss our final results.

Event selection
The choice of selection criteria is guided by two main requirements: the optimization of the signal significance (S/ √ B), and the compatibility with trigger and data acquisition constraints.The main features of the signal, to be exploited in the event selection, are: presence of two, high-p T , b jets, showing an invariant-mass peak; presence of a pair of jets in the forward and backward rapidity regions.In principle such a signal could also exhibit rapidity gaps, due to the colour-singlet exchange of EW bosons among the incoming hadrons; this fact has been used recently in [5].Because of the high luminosity (and the large number of overlapping events) required to study this final state, and because of the large emission rate for extra jets in WBF processes (see [6]), we do not feel comfortable with applying this additional constraint in our study.
The tagging of the b jets is only possible in the central region |η b | < 2.5.The efficiency of the tagging algorithm, furthermore, suggests using a p j T cut as large as possible.Since the measurement of the Higgs boson in this channel will take place only after its discovery and the determination of its mass, we can optimize the mass requirement by selecting only b pairs in a mass window centred around the known value of m H , up to the dijet mass resolution.These considerations lead to the following set of cuts: δ m being the experimental resolution ≃ 12%.Given the very small width of the Higgs boson in the mass range we shall consider (m H < 140 GeV), this last requirement reduces the signal to 68% of what obtained with perfect mass resolution.In the following we shall assume a b-tagging efficiency ǫ b = 0.5.While harder cuts on p b T would improve the S/B ratio, they would also risk sculpting the mass distribution, setting a higher value for the dijet mass threshold and therefore making it harder to extract the background shape directly from the data.
The large momentum exchange required for the emission of the space-like gauge bosons will lead to a hard p j T spectrum for the forward and backward light jets.This is clearly shown in Fig. 1 3 , where we see that the jet p T peaks at approximately 30 GeV.The spectrum of typical QCD backgrounds will viceversa peak at low p j T .The large momentum of the forward jets, and their large rapidity separation, favours large dijet invariant masses, as can be seen from Fig. 2. The cuts we select for the two jets are: The large p j T cut is driven by the requirement that trigger rates be kept at acceptable levels (see later).We present the two cases of 60 and 80 GeV to display the sensitivity to this threshold.A final choice will presumably only be possible with a complete detector simulation, or once the background data will be available.As we will comment later, the cut on p j T above 80 GeV is also very efficient in decreasing the backgrounds due to multiple overlapping events.The large mass cut is selected to reduce as much as possible the QCD jet backgrounds.This cut, in addition to the rapidity cut, is also efficient in removing the contamination from the process gg → Hgg, as shown in ref. [3].
In addition to the above cuts, we shall consider two alternative selection criteria for the light-jet rapidities, labelled (a) and (b).The case (a) is given by: 2.5 < |η j | < 5, η j 1 η j 2 < 0, (9) while for the the case (b), we only have the condition: In the case (b) we verified that requiring m jj > 1000 GeV forces the product η 1 • η 2 to be negative for the largest fraction of the events.By inspection of the differential distributions for the variable ∆R bb we find that cutting ∆R bb < 2 for the configuration (a) gives an additional enhancement of the signal with respect to the backgrounds.

The study of signal and backgrounds
The background sources we considered include: The cases with 4 light-jet events are considered since the experimental resolution leads, for any tagging algorithm, to a finite probability of b tags in light jets (fake tags).We shall label light jets mistagged as b jets with the notation j b , and assume two possible values of fake tagging efficiencies ǫ f ake , 1% and 5%.While the first choice is probably optimistic, given the presence of real secondary vertices in jets containing a charm quark, the second is likely to be too conservative.As we shall see, however, the requirement of tagging both b jets renders in any case the backgrounds with real b quarks the dominant ones.
The calculation of signal and background events is based on the numerical iterative procedure ALPHA [7], as implemented in the library of MC codes ALPGEN [6].While ALPGEN allows for the full showering of the final states, both in the case of signals and backgrounds, all our calculations are limited to the parton level.This is because a realistic estimate of the rates would anyway require a full detector simulation, which is beyond the scope of this paper.
Table 1: Signal and background events for configuration (a), with p j T > 60 GeV, for three possible values of the Higgs mass.
The jjjj entry includes the squared b−mistagging efficiency (ǫ f ake = 0.01).The first raw relative to the Z * /γ * contribution refers to the effect of the physical mass tail, while the second raw refers to the finite experimental Z mass resolution, (δm Z /m Z = 0.12).The integrated luminosity is 600 fb −1 .The PDF set used is CTEQ4L.See the text for the description of other, smaller, backgrounds.The event rates are obtained using the parametrization of parton densities CTEQ4L.Given the overall uncertainties of the background estimates, the results are not sensitive to this choice.The renormalization and factorization scales have been chosen equal (Q).In order to be conservative in the background estimates, we selected as a default for our study a rather low scale, namely Q 2 = p 2 T , where the average is taken over all light and b jets in the event 4 .In view of the large ŝ values of the elementary processes involved, due in particular to the large mass threshold for the pair of forward jets, we believe that our background rates may be overestimated by a factor of at least 2. In spite of this we prefered the conservative approach, in order to present a worse-case scenario.The backgrounds are much more sensitive to the scale choice than the signal, due to the larger power of α s .The background uncertainty will not however be a limitation to the experimental search, since the background rate should be determined directly from the data, as we shall discuss.
Tables 1-4 present our results for signal and backgrounds, for the following cases: (i) p j T > 60 GeV and rapidity configuration (a); (ii) p j T > 60 GeV and rapidity configuration (b); (iii) p j T > 80 GeV and rapidity configuration (a); (iv) p j T > 80 GeV and rapidity configuration (b).The numbers correspond to 600 fb −1 of integrated luminosity, namely the expected value for three years of running of ATLAS and CMS with an instantaneous luminosity of 10 34 cm −2 sec −1 .The numbers relative to final states with mistagged jets include the square of the mistagging probability ǫ f ake = 0.01.
We shall now discuss each individual background contribution in detail.

Single-interaction events
The 4-jet backgrounds originating from a single hard collision are shown in the second and third rows of Tables 1-4.In the case of the j b j b jj background, we accept all events in which at least one pair of light jets passes the cuts in eqs.( 1)-( 4), and the other two jets satisfy eqs.( 5)-( 8), in addition to the appropriate rapidity cut (eq.(9) or ( 10)).As anticipated, the contribution from real b jets is the dominant one, even assuming ǫ f ake = 0.05.From the numbers in the Tables 5 and 6, we see that the S/ √ B can be as large as 5.However, the ratio S/B is only a fraction of a percent.This implies that the background itself will have to be known with accuracies at the permille level.There is no way that this precision can be obtained from theoretical calculations.The background should therefore be determined entirely from the data.We expect our kinematical thresholds to be low enough not to sculpt the shape of the b b mass distribution at masses close to the Higgs mass.This is true for the leading 4 jet backgrounds, as shown in Fig. 3.The b b invariant mass of the simulated b bjj background is shown here to be well behaved in the [100, 150] GeV region.The distribution in the case of the j b j b jj final states is similar.As a result, we expect that the sidebands of the Higgs signal (the regions of mass below m H (1 − δ m ) and above m H (1 + δ m )) can be safely interpolated in the region under the Higgs peak, similarly to what was done by UA2 in the extraction of the W/Z → jj decay [8].
For this extraction to be possible, however, full background samples have to be collected.The large rate of untagged jjjj events could therefore give problems with the triggers and with the data acquisition.This is because the b tagging algorithm is typically applied only offline, and therefore a number of untagged jjjj events larger than what is acceptable by the trigger and by the data acquisition would force higher cuts, or a trigger prescaling, strongly reducing the number of recorded signal events.Removing the fake-tagging probability from the numbers in the Tables 1-4, leaves untagged jjjj rates in the range of few×10 7 and 10 9 , depending on whether configuration (a) or (b) is chosen.Since the mass window for the signal is approximately 30 GeV wide, these rates must be increased by a factor of 3-4, to allow for a sufficient coverage of the sidebands of the b b mass distribution, coverage which is required to enable the interpolation of the background rate under the Higgs mass peak.The numbers in the Tables 1-4 refer to 6 years of data taking, corresponding to 6 × 10 7 s, distributed among the two experiments.The result is a rate of events to tape in the range of 1 Hz (for configuration (a) with 80 GeV jet threshold) up to 50 Hz (for configuration (b) with 60 GeV jet threshold).While a 1 Hz rate to tape is acceptable, 50 Hz would almost saturate the expected data acquisition capability of 100 Hz.In this last case, some extra information would have to be brought into the trigger.The best candidate is some crude b-tagging.If a rejection against non-b jets at the level of 20% per jet could be achieved at the trigger level, the rates would be reduced by a factor of 20, down to perfectly acceptable levels.While the above processes represent the largest contribution to the backgrounds, the smoothness of their mass distribution in the signal region allows to estimate their size with statistical accuracy, without significant systematic uncertainties.The situation is potentially different in the case of the backgrounds from the tails of the Z decays.The Z mass peak is sufficiently close to m H , especially in the case of the lowest masses allowed by current limits, to possibly distort the m bb spectrum and spoil the ability to accurately reconstruct the noise level from the data.The size of the two possible effects (smearing induced by the finite experimental energy resolution and the intrinsic tail of the Drell-Yan spectrum) are given in the 4th and 5th rows of the Tables 1-4.Aside from the case of the largest m H value, where these backgrounds are anyway negligible, the dominant effect is given by the detector resolution.For the configurations (a) these backgrounds represent a fraction of the order of at most 40% of the signal, at small m H , rapidly decreasing at higher m H .For the configuration (b), the rates are comparable to the signal at low m H .A 10% determination of these final states, which should be easily achievable using the (Z → ℓ + ℓ − )jj control sample and folding in the detector energy resolution for jets, should therefore be sufficient to fix these background levels with the required accuracy.As for the contribution of the on-peak (Z → b b)jj events to the determination of the sideband rates, we verified that their impact is negligible.We obtain a number of the order of 60K events with 600 fb −1 in the mass range 83-100 GeV, for configuration (b) and p T > 80 GeV for the forward jets.These events can therefore be subtracted from the sidebands with a statistical accuracy better than 1% using the measurement of the on-peak (Z → ℓ + ℓ − )jj final states.It should be pointed out that extrapolating from the leptonic to the b b rates with this accuracy requires a matching precision in the knowledge of the tagging efficiencies, something which remains to be proven.
Before concluding the list of single-interaction backgrounds, we briefly comment on the smaller contributions, pp → t t and pp → t tj, with t decaying hadronically.Before applying the cuts, we adopt a clustering algorithm for the jets coming from the decay of a W .We sum the four-momenta every time the separation between the two jets is below the threshold ∆R = 0.4.This happens quite often, since in order to have a pair of jets in the event with an invariant mass above 1 TeV at least one of the two W s coming from the t decays must have a large boost.After this clustering algorithm, using the event selection (b), about 300 t tj events survive the cuts at 600 fb −1 , while the number of t t events is negligible.The configuration (a) leads to even smaller rates.The absolute rate can be fixed using the data, by reconstructing the individual tops.This should be particularly simple, since the request of large dijet mass forces the t and t to be very well separated, and the large momentum of the W 's will reduce the combinatorial background in the association of the b jets with the W jets.

Overlapping events
We come now to the study of events due to the superposition of multiple pp interactions.The reason why these events are a potential problem is that while production of large dijet invariant masses in individual events is strongly suppressed energetically, these can accidentally appear when mixing jets produced in separate events (after all the overall energy available in 2 collisions is twice that for a single pp collisions): for example, we can consider two events, one in which a small-mass dijet pair is produced with large positive rapidity, the other in which a low-mass pair is produced at large negative rapidity; the pairing of jets from the two events will lead to large rapidity separations, and to large dijet masses.
In the simplest case of two overlapping events, we have four possible combinations of events leading to a b bjj background: , where (ab) ≡ pp → ab.Since we do not veto on the presence of extra jets, triple events such as (j 1 j b ) ⊕ (jj 2 ) ⊕ (j b j) are also possible.The probability of having n simultaneous events with a jj final state during a bunch crossing, assuming a bunch crossing frequency of (25 ns) −1 , is given by the Poisson probability distribution function π n (µ) with average µ = 0.25 × σ(pp → jj)/mbarn × L/L 0 , where L is the instantaneous luminosity and L 0 = 10 34 cm −2 sec −1 .
To estimate the rates, we first generate a sample of unweighted events of the type pp → jj.We then randomly extract from this sample n-tuples of dijet events, which are associated to events where n dijet pairs from n proton-proton collisions are created in the same bunch crossing.The background can be then estimated as: where B is the number of bunch crossings accumulated during the run time, and p n = f n /N n , (n = 2, 3), where N n is the total number of n-tuple events generated, f 2 , f 3 are the number of double and triple events passing the selection cuts found in the sample of generated events.Ellipses denote simultaneous collisions of higher order.Since π n (µ) drops quite rapidly with increasing n, we limit our analysis at n = 3.The above formula can be easily modified to include the presence of σ(pp → b b) events.All numbers given below refer to the case of high luminosity, namely 10 34 cm −2 s −1 .Since these rates scale quadratically, they should be reduced by a factor of 100 in the case of 10 33 cm −2 s −1 .We verified that the most dangerous background comes from events of the type (jj b ) ⊕ (jj b ).The main reason is as follows: since the forward, non-tagged jets are required to have a large p T threshold (60 or 80 GeV), the fake b jets in the central region will inherit the same transverse momentum cut, as they are produced back-to-back with the related forward jet.As a result, the invariant mass spectrum of the j b j b pair will have a shape peaked at about twice the cut, and therefore right in the middle of the signal region.Typical shapes of the m bb spectra are given in Fig. 4, for configuration (a) (The shapes for configuration (b) are very similar).In the case of 60 GeV, the signal regions are right in the middle of the background peak, or on its rising slope; this makes the background estimate very sensitive to the assumed energy resolution, both in the forward region (since the energy scale in the forward region affects the onset of the trigger for the forward jets, thus affecting the spectra of the central jets recoiling against them) and in the central region as well (since the mass spectrum is rapidly rising in the 100-150 GeV range.Our results were obtained by assuming a forward jet energy resolution given by σ f wd = √ E ⊕ 0.07 E, in addition to the 12% mass resolution used earlier for the central jets.The distributions in Fig. 4 include this resolution smearing.The rates obtained after including the resolution effects are approximately twice as large as those obtained with perfect resolution, stressing the importance of these effects.In absolute terms, the Tables 1-4 show that these contributions are of the same order of magnitude as the signal when p j T > 60 GeV is used, but much smaller when the higher p j T thereshold is used.In the former case, these final states are a potential threat, unless a way can be found to estimate from the data their exact size.This cannot be done using the mass spectrum in the sideband regions, since the rate is too small compared to the leading 4-jet processes.We believe that it should be possible however to use the distribution of the z vertex separation between the two events as a diagnostic tool.Since the two tagged jets come from different pp events, and given that the spread of the interaction point in z is of the order of few cm, the fraction of overlapping events where the z positions of the two vertices cannot be separated should be of the order of 10%, a number measurable by extrapolating the ∆z distribution from large values, down to the range in which ∆z is of the order of the experimental resolution. Other sources of backgrounds from overlapping events are less dangerous.Events where the b b or j b j b pair comes from the same hard interaction ((b b)⊕(jj) and (j b j b )⊕(jj)) have a smooth mass spectrum in the 100-150 GeV region, and rates smaller than those of the single-interaction b bjj or j b j b jj events.The mass spectrum of (b b) ⊕ (jj) events is shown in Fig. 3 5 .Their contribution can therefore be estimated precisely from the data 6 .In the specific case of m H = 120 GeV, for example, we obtain the following numbers of events: 10 T > 80 GeV cut.The events from three separate pp collisions contribute less than 10% of the twocollision rates shown in the Tables 1-4, at 10 34 cm −2 s −1 .

Results
Tables 5-8 summarize our results for the sensitivity defined as the ratio of the number of signal events divided by the square root of the number of background events for different values of the mistagging efficiency ǫ f ake .Tables 9,10 show our results on the determination of the branching ratio B(H → b b) and accordingly on the Hb b Yukawa coupling y Hbb , Table 5: The sensitivity, defined as the ratio of the number of signal events divided by the square root of the number of the background events.The mistagging efficiency of light jets, ǫ f ake , is ǫ f ake = 0.01.The integrated luminosity is 600 fb −1 for both configurations (a),(b), and the transverse momentum cut on jets is p j T > 60 GeV.assuming the knowledge of the HW W coupling.This can be determined using other channels, as discussed in the literature [9].These results rely also on the assumption of SU(2) invariance to relate the contributions to the signal coming from the HW W and HZZ couplings, which can not be experimentally disentangled in the WBF production mechanism.With a total luminosity of 600 fb −1 , a relative precision of about 20% on the B(H → b b) branching ratio can be attained.This represents an improvement with respect to what obtained in other channels [10,11].As for the Hb b Yukawa coupling, a statistical significance of at best 30% is reachable7 .The significance is rather flat in the 115-140 GeV mass range, as a result of the compensation between overall rate (which decreases at larger masses) and sensitivity of the BR to the Yukawa coupling (sensitivity which increases at smaller BR, for larger masses).The effect of applying a larger cut (80 GeV) on the transverse momentum of forward jets is to reduce by approximately 10% the statistical accuracy of the measurement.This choice could however turn out to be more reasonable in view of the reduced experimental difficulties at larger p j T .The H → b b decay in the WBF channel also allows for a model independent determination of the ratio of widths Γ(H → b b)/Γ(H → τ + τ − ) when combined with the qq → qq(H → τ + τ − ) mode [12].This determination can be compared with what obtained in the t tH production channel by [11].Moreover, comparing the WBF mechanism studied in this paper with the associated W (H → b b) production, one could test the SU(2) relation between the SM HW W and HZZ couplings for low Higgs masses.

Conclusions
In this letter we examined (H → b b)jj production at the LHC, with the goal of assessing the potential accuracy in the determination of the y Hbb Yukawa coupling.A study of the observability of this channel has also been presented in ref. [5].We believe our paper provides a more realistic evaluation of the experimental challenges of this measurement, and find less optimistic results.
In particular, we identified two main sources of backgrounds: • 4 jet final states: these are over 100 times larger than the signal, but could be evaluated with accuracy using the sidebands of the b b mass spectrum.This requires however some tagging information to be available at the trigger level, to reduce to acceptable levels the data storage needs for inclusive, untagged, 4 jet final states.
• 4 jet final states from multiple collisions: a large contribution comes from events of the type (jj b ) ⊕ (jj b ), where the b b mass spectrum has a broad peak in the middle of the signal region.The absolute rate of these events (of the order of the signal rate, when using the lower transverse momentum threshold of 60 GeV) can be determined if the distribution of the z vertex separation between the two overlapping events can be determined with a resolution of the order of 5-10mm.These events are significantly reduced in number when using the higher threshold of 80 GeV for the forward jets.
Our parton-level analysis should be completed with a full detector simulation, but, already at this stage, it provides a strong indication for the relevance of this channel for the B(H → b b) branching ratio.We have shown in fact that the B(H → b b) can be measured with a 20% precision for an Higgs mass around 120 GeV assuming that the coupling HW W is the one predicted by the Standard Model or determined in other reactions already studied in the literature.We also observe that the WBF channel we study, combined with other processes, can be used for a model independent determination of the To conclude, we should point out that all statistical accuracies listed in this study should be matched by an excellent control over experimental systematics, including the knowledge of b-tagging efficiencies (needed for example to allow the determination of Z → b b backgrounds from the measurement of Z → ℓ + ℓ − final states) and their dependence on the b momentum, and of forward jet tagging efficiencies and fake (pile-up or calorimeter noise) rates.On the other hand, as mentioned at the beginning, we expect our estimates of the physics backgrounds to be very conservative, being based on very low Q 2 scales for the evaluation of the strong coupling constant; furthermore, we anticipate that more sophisticated analyses based on kinematical correlations in the event (exploiting for example the scalar nature of the Hb b coupling) will help improving the signal significance.

Figure 1 :
Figure 1: The p j T distributions are shown: high p j T regions are more suppressed in the b bjj QCD background (solid) with respect to the signal (dashes).The inclusive distributions shown are normalised to the same cross section.

Figure 2 :
Figure 2: The distribution for m jj is shown both for the signal (dashes) and for the b bjj QCD background (solid).The inclusive distributions shown are normalised to the same cross section.

1 . 4 . t t production 5 . t tj production 6 .
QCD production of b bjj final states, where j indicates a jet originating from a light quark (u, d, s, c) or a gluon; 2. QCD production of jjjj final states.3. Associated production of Z * /γ * → b b and light jets, where the invariant mass of the b b pair is in the Higgs signal region either because of imperfect mass resolution, or because of the high-mass tail of the intermediate vector boson.b bjj and jjjj production via overlapping events.

Figure 3 :
Figure 3: The distribution of the invariant mass of the system b b in the b bjj QCD background (solid line), and in overlapping events of the type (b b) ⊕ (jj) (dashed line).The curves are normalised to the same cross section.

Figure 4 :
Figure 4: The distribution of the invariant mass of the system b b in the j b j ⊕ j b j multiple-collision QCD background, for configuration (a).

5 and 4 ×
10 5 (jj) ⊕ (b b) events for p j T > 60 GeV in the configurations (a) and (b), respectively; 6 × 10 4 and 2 × 10 5 (jj) ⊕ (b b) events for p j T > 80 GeV in the configurations (a) and (b), respectively.The contributions from (jj) ⊕ (j b j b ) final state are smaller by a factor of approximately 12, independently of the configuration and transverse momentum thresholds, and assuming ǫ f ake = 0.01.Events of the kind pp → b b ⊕ pp → b b turn out to be totally negligible, at the level of 40 with the p j

Table 3 :
Same as Table 1, with p j T > 80 GeV.

Table 6 :
The same as Table5, with p j T > 80 GeV.

Table 7 :
The same as Table5but with a mistagging efficiency of ǫ f ake = 0.05.

Table 8 :
The same as Table6but with a mistagging efficiency of ǫ f ake = 0.05.

Table 9 :
The statistical significance of the determination of the branching ratio Γ b /Γ and of the b-quark Yukawa coupling in the configurations (a) and (b).A luminosity of 600 fb −1 is assumed; the transverse momentum cut on jets is p j T > 60 GeV.Here ǫ f ake = 0.01.Using ǫ f ake = 0.05 will worsen these estimates by approximately 20%.

Table 10 :
The same as Table9with p j T > 80 GeV.Hbb /y Hτ τ ratio and for a test of the ratio of the couplings g HW W /g ZW W for low Higgs masses.