INTRODUCTION

Onions (Allium cepa L.) are very sensitive to interfer ence from weeds. Onion cultivation is characterized by low soil cover because of the low plant size, little production of leaves, which are cylindrical and erect, and slow initial growth of the crops. These aspects result in low competitiveness and allow the germina tion and growth of weeds at practically any stage of development (^{Reis et al., 2017}).

In terms of crops with a high technological level, this crop presents a high cost of production; it is common to find high levels of fertility in this production sys tem, with systematic irrigation and intensive tillage, which promote high populations of weeds (^{Pereira, 2008}). The management of weeds in onion crops is an important variable in the definition of produc tivity since this crop is negatively affected by com petition with weeds. Onion productivity decreases drastically with the interference of weeds, which may reach 100% loss in the commercial production of bulbs (^{Jean-Simon et al., 2012}; ^{Qasem, 2005}). Thus, it is necessary to use control measures during the criti cal period of interference prevention, which is long in the onion crop and varies from 40 to 100 d (^{Soares et al., 2003}; ^{Qasem, 2006}).

Among the control methods, chemical management of weeds is widely used in this crop because of greater yield in large areas, high efficiency and lower cost. However, it requires knowledge on the use and be havior of herbicides in the soil and their selectivity to the crop, especially because the onion is highly sen sitive to these products (^{Qasem, 2006}; ^{Uygur et al., 2010}; ^{Carvalho et al., 2014}).

In Brazil, there are few herbicide options for onion use. Linuron has been the standard for chemical weed control in onion crops since the 1980s. However, the control exerted by this herbicide has been unsatisfac tory because, after 10-15 d, post-emergence herbi cide applications, which may toxify the crop, and/ or manual weeding are required, increasing the cost of production.

Flumioxazin (N-[7-fluoro-3,4-dihydro-3-oxo-4-prop-2-ynyl-2H -1,4-benzoxazin-6-yl]cyclohex-1-ene-1,2-dicarboxamide) may be an alternative for herbicide use in onion crops. It is an active non-ionic ingredient that exhibits low solubility in water (1.79 mg L^-1 at 25°C) (^{Ferrell et al., 2005}) and a vapor pres sure of 2.41x10 ^s mm Hg at 22°C, indicating a low vol atilization potential (^{Rodrigues and Almeida, 2011}). It may present half-life of 17.6 d in the soil (PPDB, 2018) and undergoes degradation through hydroly sis and microorganism activity (^{Alister et al., 2008}). Accordingly, flumioxazin presents little aggression to the environment; however, there are reports that, in some areas of Alliaceas in Alto Paranaíba, damage occurred through phytotoxicity, mainly with excess rainwater or irrigation and the use of high doses or doses at an inappropriate vegetative stage, making farmers afraid to use it.

Flumioxazin is registered for pre-emergence applica tions on weeds, 2-3 d after seedling transplanting (^{Agrofit, 2018}). However, in the direct sowing sys tem, applications on onion plants at the true four-leaf stage, the recommended stage for most herbicides, is considered late and ineffective against the advanced development of weeds.

Because of these problems, an alternative is to use reduced doses at the initial stages of the crop (1^st, 2^nd, and 1^st + 3^rd true leaf) since recommended doses for the onion vary from 60 to 90 g ha^-1 ai. The ob jective of this study was to evaluate the efficacy of reduced doses of flumioxazin, applied in the initial phenological stages of the onion crop in terms of bulb productivity.

MATERIAL AND METHODS

Two experiments were carried out under field con ditions. Experiment 1 was set in the experimental area of Cooperativa Agropecuária do Alto Paranaíba (COOPADAP), located in the municipality of Rio Pa ranaíba, Minas Gerais, Brazil, in a clayey Dystroferric Red Oxisol from February to July of 2015. Experiment 2 was installed in an area of commercial production on the Santo Amaro Farm, located in the municipal ity of Rio Paranaíba, in a dystrophic Red Oxisol with a clay texture from February to June of 2016.

In both experiments, direct sowing was used with four double lines per bed, where one million and two hundred thousand seeds per hectare were distributed. In experiment 1, cv. Perfecta was sown and, in experi ment 2, cv. Sirius was used.

The planting fertilization in experiment 1 was 800 kg ha^-1 of P₂O₅, 50 kg ha^-1 of N and 50 kg ha^-1 of K₂O. At 20 and 40 d after emergence, topdressing fertilization was carried out with 75 kg ha^-1 of N and 125 kg ha^-1 of K₂O. In experiment 2, 950 kg ha^-1 of P₂O₅, 40 kg ha^-1 of N and 30 kg ha^-1 of K₂O were used. At 25 and 50 d after emergence, topdressing fertilization was carried out with 55 kg ha^-1 of N and 130 kg ha^-1 of K₂O. The fertilization differences between the areas were based on the chemical analyses of the soils and nutrition demand of the cultivars.

In both experiments, a randomized complete block design was used, with treatments arranged in a 5x3+1 factorial scheme, with four replicates. The first factor corresponded to the reduced doses of flumioxazin (5, 10, 15, 20 and 25 g ha^-1), and the second factor was the phenological stage of the onion on which the herbicide was applied (1^st, 2^nd, and 1^st + 3^rd true leaf), as well as weed control. The parcels with the treatment 1^st+3^rd true leaf received two applica tions of flumioxazin in the two phenological stage of the onion at the five doses described.

The dimensions of each experimental plot were 1.5 (four double lines) x 4 m, and the two central double lines were considered as the useful area, minus 0.5 m at each end of the plot. For the application of flumioxazin (Flumyzin 500 - active ingredient concen tration 500 g kg^-1 - wettable powder formulation), a CO₂ backpack sprayer pressurized at 2 bar equipped with a bar with three flat jet tips, (fan type) 110.02, spaced at 0.5 m was used at 0.5 m above the soil with a volume of 200 L ha^-1.

At 15, 30 and 45 d after application (DAA) of the her bicide, counted from the application on the 1^st true leaf stage, phytotoxicity evaluations were performed using a visual scale from 0 to 100%, where 0% repre sents the absence of symptoms and 100% is the death of the plant, and visual evaluation of the weed con trol (^{Gazziero, 1995}). At 30 and 60 DAA, stand evalu ations were performed by counting plants in two linear meters at three delimited points in each plot.

At 150 and 130 d after sowing, the onion was har vested in Experiments 1 and 2, respectively. One hundred plants were harvested in the useful area of each plot, and the commercial (Classes 2, 3, 4, and 5) and non-commercial (Classes 0 and 1) and total bulb yields were classified and estimated. The classifica tion was performed according to the Companhia de Entrepostos e Armazéns Gerais de São Paulo (CEAG-ESP) commercial scale, based on the equatorial diam eter of the bulbs in Class 0 (<1.5 cm), Class 1 (1.5 to 3.5 cm), Class 2 (3.5 to 5.0 cm), Class 3 (5.0 to 7.0 cm), Class 4 (7.0 to 9.0 cm), and Class 5 (>9.0 cm).

The data were submitted to analysis of variance by the F test (P<0.05), and the means were plotted in figures with standard error. The experiments were analyzed separately. The software Sisvar 5.6 and Sigmaplot 10.0 were used for this.

RESULTS AND DISCUSSION

The onion toxicity was affected by the phenological stage of the crop, reduced doses of flumioxazin and cultivar. In experiment 1, at 15 DAA, the plants pre sented 15% injury on average (Fig. 1A), character ized by chlorotic spots on the leaves. In experiment 2, the application performed on the 2^nd leaf caused increased phytotoxicity as the doses increased up to a maximum of 25% (Fig. 1D). At the time of this evaluation, the application on the 3^rd leaf stage had not yet been performed because the onion plants still had two true leaves, explaining the equal results with the applications performed in the stages 1^st and 1^st + 3^rd leaf.

Figure 1 Phytotoxicity of onion plants evaluated at 15, 30 and 45 d after application of reduced doses of flumioxazin in experiment 1 (A, B, C) and experiment 2 (D, E, F), respectively. 

According to ^{Cavaliere (2015)}, in order to minimize the toxicity caused by herbicides in onions, fractionation and escalation of doses have been carried out and have contributed to increased selectivity in onions, especially in the initial stages of growth of direct sowing plants. However, even with the reduc tion of doses, phytotoxicity may occur, as verified in the present study. In Turkey, onion phytotoxicity from herbicides is also a serious problem and, even when splitting the recommended dose of oxyfluorfen, injures have been found in plants (^{Uygur et al., 2010}).

In the visual evaluations performed at 30 and 45 DAA in experiment 1, higher phytotoxicity was observed for the plants submitted to application on the 1^st + 3^rd leaf, followed by application on the 2^nd leaf and on the 1^st true leaf of the crop (Fig. 1B and 1C). In ex periment 2 at 30 DAA, the application on the 1^st+3^rd leaf caused greater phytotoxicity in the plants than at other stages, with reduction of the symptoms at 45 DAA, regardless of the treatments (Fig. 1E and 1F).

The increase in phytotoxicity in relation to the phenological stages of the crop may evidence that the tolerance of the onion to this herbicide decreases with the age of the plant and the number of applica tions (1^st + 3^rd leaf). This is probably related to the in crease in the number of leaves and, consequently, the leaf area for interception of the product at those ini tial stages of the onion planted via direct sowing. For oxyfluorfen, a herbicide of another chemical group, but with the same mechanism of action as flumioxazin, studies have shown that symptoms decreased with the age of the transplanted onion, which is due to the increase in the wax content in the leaves (^{Carvalho et al., 2014}). These authors verified greater phytotoxicity when oxyfluorfen was applied at 14 d after transplanting (DAT) when compared to the ap plication at 28 DAT.

At 60 DAA, the onion plants completely recovered from toxicity, regardless of the phenological stage (data not shown). This recovery was due to the low or absence of translocation of the herbicide in the plant, acting as a contact product. Thus, the new emitted leaves were free of toxicity symptoms. ^{Durigan et al. (2005)} found that, at 49 DAA of flumioxazin, onion plants recovered from severe initial toxicity when the application was made on the 5^th-6^th leaf of the trans planted crop.

Weed control in both experiments at 15 DAA was sat isfactory in the plots where application was made on the 1^st leaf (Fig. 2A and 2D). The application on the first true leaf of the onion provided weed control at an early stage of development, ensuring control ef fectiveness even at the smallest evaluated dose. The predominant species in experiment 1 were Nicandra physalodes (L.) Gaertn., Galinsoga parviflora Cav., Portulaca oleracea L., Sonchus oleraceus L., and Ageratum conyzoides L. In experiment 2, the predominant weeds were Eleusine indica (L.) Gaertn, Commelina benghalensis L., N. physalodes and Gamochaeta coarctata (Willd.) Kerguélen.

Figure 2 Weed control evaluated at 15, 30 and 45 d after the application of reduced doses of flumioxazin in experiment 1 (A, B, C) and experiment 2 (D, E, F), respectively. 

The application on the second leaf of the onion satis factorily controlled only with the dose of 10 g ha^-1 in experiment 1 at 15 DAA (Fig. 2A). In the other evalua tions (30 and 45 DAA), this treatment was ineffective (Fig. 2B and 2C). The application of reduced doses of flumioxazin when the crop had two true leaves was not effective probably because the weeds in the area were already more developed. In experiment 2, this application stage provided acceptable control from 30 DAA, regardless of the applied dose (Fig. 2E and 2F).

At 45 DAA in experiment 1, the applications on the 1^st and 1^st+3^rd leaves maintained control of over 90% with the 10 g ha^-1 dose of flumioxazin (Fig. 2C). Be sides controlling weeds early in the area when applied at first leaf of the onion, the additional application on the third leaf extended the residual effect of the her bicide, which also had a pre-emergence action on the weeds. In experiment 2, the application in the three stages provided satisfactory control (>80%) (^{Gazziero, 1995}), particularly on the 1^st and 3^rd leaves, which maintained control near 100% regardless of the dose (Fig. 2F).

In experiment 1, the stand was not altered by the ap plication of reduced doses of flumioxazin in relation to the weeding control in the two evaluated periods (Fig. 3A and 3B). In experiment 2, the stand was re duced only with the highest dose (25 g ha^-1) at 30 DAA in the three application stages and, at 60 DAA, in the 1^st and 1^st + 3^rd leaf stages (Fig. 3C and 3D). The reduction in the number of onion plants was also verified by ^{Ferreira et al. (2000)} with an application of oxyfluorfen (0.192 kg ha^-1) at 19 d after sowing, reducing the initial stand by 22.6%.

Figure 3 Onion stand evaluated at 30 and 60 d after the application of reduced doses of flumioxazin in experiment 1 (A, B) and experiment 2 (C, D), respectively. 

Because the onion crop is very sensitive to herbi cides, it is likely to suffer toxicity and be suppressed, especially in the early stages of development. Thus, depending on the dose and growth stage of the crop, phytotoxicity may be severe, irreversibly im pairing the growth of the plants, leading to death and, therefore, reducing the initial population of the crop. Onion production fields established by di rect sowing usually present a highly variable plant population, which was also observed in the experi mental plots.

The stand verified in the two experiments was dif ferent in the absence of herbicide (Fig. 3), which con tributed to the difference in bulb yield observed in the two experiments (Fig. 4). The lowest number of plants per area reduced intraspecific competition, re sulting in bulbs with a larger diameter and weight.

Figure 4 Productivity of non-commercial, commercial and total bulbs of onion plants submitted to reduced doses of flumioxazin at different phenological stages in experiment 1 (A, B, C) and experiment 2 (D, E, F), respectively. 

The application of flumioxazin on the 2^nd and 1^st+3^rd leaves increased, in most doses, the bulb yield of 'Perfecta' non-commercial bulbs (Fig. 4A). The application at the first leaf was equal to or lower than the control, up to 15 g ha^-1, increasing the productiv ity of bulbs with a diameter less than 3.5 cm in the two highest doses. In experiment 2, the productivity of non-commercial bulbs was similar in the absence and presence of reduced doses of flumioxazin, except for the 20 g ha^-1 dose applied at the 1^st true leaf of the crop, which was lower than the control without an application (Fig. 4 D). Intoxication caused by herbi cides may result in reduced crop growth and reduced quantity and quality of the products harvested (^{Rob inson, 2008}), even with the application of reduced doses.

The commercial and total yields in experiment 1 presented the same behavior, with emphasis on the application in the first leaf, which was superior to the other phenological stages up to the dose of 20 g ha^-1. The dose of 25 g ha^-1 of flumioxazin applied on the 1^st+3^rd leaf reached productivity similar to the first leaf stage. However, in this experiment, the application of flumioxazin reduced the total and commercial productivity by at least 10 t ha^-1, as compared to the absence of herbicide (Fig. 4B and 4C). The cultivars presented different tolerances to flumioxazin, where 'Perfecta' - experiment 1 was less tolerant than 'Sirius' - Experiment 2 (Fig. 1). This fact contributed to the effect of the herbicide on the reduction of commercial and total bulb pro ductivity and to the increase of the productivity of non-commercial bulbs, in relation to the weeding control in experiment 1.

This differential tolerance can be attributed to dif ferent foliar waxiness (^{Ferreira and Costa, 1982}) in the two cultivars, which needs to be investigated. According to the technical information from the seed company, both cultivars have good waxiness; however, it is necessary to evaluate the thickness and composition of the cuticle and the amount of wax present in the leaves of the two cultivars to better explain the influence of these characteristics on the differential tolerance to flumioxazin and others her bicides applied post-emergence.

In experiment 2, starting with the 15 g ha^-1 dose, a difference was found in commercial and total productivity between the stages, with superiority found in the application on the first leaf in com parison to the others. For the control without her bicides, the application on the 2^nd leaf reduced the commercial and total productivity, the application on the 1^st+3^rd leaf was similar in all the doses and the application on the first leaf was similar up to the dose of 20 g ha^-1 and higher at the dose of 25 g ha^-1 (Fig. 4E and 4F). In this experiment, it was evi denced that the use of reduced doses of flumioxazin may guarantee productivity equal to or even higher than the weeding control, depending on the dose applied and the phenological stage of the crop. In Iran, the use of 75% of the recommended dose of oxyfluorfen provided quality and bulb production comparable to the absence of herbicide according to a study by ^{Abbaszadeh et al. (2014)}.

The results evidence the efficacy of reduced doses of flumioxazin in onion crops at early stages (1^st, 2^nd and 3^rd true leaf). The dose of 20 g ha^-1 showed use potential in the two experiments for 'Perfecta' and 'Sirius', enabling a reduction of 77% to 88% of the commercial dose recommended for onions planted with seedling transplanting. In addition, because of the lack of information in the literature on the be havior of flumioxazin in soil in tropical conditions, this reduction in the dose reduces the potential risks of environmental contamination.

CONCLUSIONS

The application of flumioxazin on the 2^nd leaf re duced commercial productivity and was ineffective at controlling weeds. The application on the 1^st+3^rd leaf, although effective in the control, caused greater phytotoxicity and, therefore, reduced commercial productivity. The best strategy for weed manage ment was 20 g ha^-1 of flumioxazin applied to the on ion plants when they reached the true first leaf stage